Using Ccproxy Inspector
starbaser/ccproxy
Operates the ccproxy inspector MITM system for intercepting, inspecting, and transforming LLM API traffic.
Answer questions about Tlon, Urbit, Tlon Messenger, Tlonbot, and OpenClaw — what they are, how they work, and how to use them.
$ npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install tloncorp/tlon-apps tlon-product-guide --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/tloncorp/tlon-apps.git skills-src && mkdir -p .claude/skills && cp -r skills-src/packages/openclaw/skills/tlon-product-guide .claude/skills/tlon-product-guide && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "tlon-product-guide" agent skill from https://github.com/tloncorp/tlon-apps/tree/develop/packages/openclaw/skills/tlon-product-guide into .claude/skills/tlon-product-guide/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "tlon-product-guide", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/tloncorp/tlon-apps/tree/develop/packages/openclaw/skills/tlon-product-guideType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install tloncorp/tlon-apps tlon-product-guide --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/tloncorp/tlon-apps.git skills-src && mkdir -p .agents/skills && cp -r skills-src/packages/openclaw/skills/tlon-product-guide .agents/skills/tlon-product-guide && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "tlon-product-guide" agent skill from https://github.com/tloncorp/tlon-apps/tree/develop/packages/openclaw/skills/tlon-product-guide into .agents/skills/tlon-product-guide/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "tlon-product-guide", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install tloncorp/tlon-apps tlon-product-guide --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/tloncorp/tlon-apps.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/packages/openclaw/skills/tlon-product-guide .cursor/skills/tlon-product-guide && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "tlon-product-guide" agent skill from https://github.com/tloncorp/tlon-apps/tree/develop/packages/openclaw/skills/tlon-product-guide into .cursor/skills/tlon-product-guide/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "tlon-product-guide", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/tloncorp/tlon-apps.git --path packages/openclaw/skills/tlon-product-guide--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install tloncorp/tlon-apps tlon-product-guide --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/tloncorp/tlon-apps.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/packages/openclaw/skills/tlon-product-guide .gemini/skills/tlon-product-guide && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "tlon-product-guide" agent skill from https://github.com/tloncorp/tlon-apps/tree/develop/packages/openclaw/skills/tlon-product-guide into .gemini/skills/tlon-product-guide/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "tlon-product-guide", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install tloncorp/tlon-apps tlon-product-guideInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/tloncorp/tlon-apps.git skills-src && mkdir -p .github/skills && cp -r skills-src/packages/openclaw/skills/tlon-product-guide .github/skills/tlon-product-guide && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "tlon-product-guide" agent skill from https://github.com/tloncorp/tlon-apps/tree/develop/packages/openclaw/skills/tlon-product-guide into .github/skills/tlon-product-guide/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "tlon-product-guide", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install tloncorp/tlon-apps tlon-product-guide --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/tloncorp/tlon-apps.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/packages/openclaw/skills/tlon-product-guide .opencode/skills/tlon-product-guide && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "tlon-product-guide" agent skill from https://github.com/tloncorp/tlon-apps/tree/develop/packages/openclaw/skills/tlon-product-guide into .opencode/skills/tlon-product-guide/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "tlon-product-guide", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
tlon-product-guideAnswer questions about Tlon, Urbit, Tlon Messenger, Tlonbot, and OpenClaw — what they are, how they work, and how to use them.
Tlon Product Guide is an agent skill from tloncorp/tlon-apps. Answer questions about Tlon, Urbit, Tlon Messenger, Tlonbot, and OpenClaw — what they are, how they work, and how to use them. Covers signup and onboarding, contacts and invites, groups (which the app calls workspaces), channels (Chat/Notebook/Bulletin/Gallery), roles and permissions, DMs, bot setup, crons, connected services (MCP), slash commands, models and API keys, privacy and encryption, hosting, exporting, self-hosting, and support. A hosted Tlonbot can use models included with a ChatGPT subscription…
Its SKILL.md is about 11k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.
It sits in Agent Workflows, covering Messaging and chat bots, Model routing and gateways and Hooks and plugins. It works with Model Context Protocol, OpenAI and OpenRouter. The repository describes itself as: Native, web clients, and server for Tlon Messenger. The licence is MIT.
12 steps, taken from the step headings in SKILL.md.
Read from SKILL.md and the folder at commit d89f6e2. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Shell commands in SKILL.md call:
justFrom the folder's file list and the shell code blocks in SKILL.md.
Links to these hosts (documentation or services it may open):
tlon.networkFrom URLs in SKILL.md, links to its own repository left out.
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Tlon Product Guide loads about 11k tokens when it runs. Until then it costs about 194 tokens; SKILL.md has 7,189 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
The full file from tloncorp/tlon-apps at commit d89f6e2, republished under its MIT licence (© tloncorp). 7,189 words, ~10,931 tokens.
.claude/skills/tlon-product-guide/SKILL.md (or your agent's skills folder).This is the reference for answering questions about Tlon Messenger and Tlonbot, walking people through features step by step, and suggesting ways to get more out of the product.
New Workspace makes one with your bot in it. Desktop's sidebar still says Home. They're the same thing, so use whichever word the person used. Someone asking for a workspace wants a group. The word means the bot's own workspace directory only when they're plainly talking about its files: asking what's in it, naming one like SOUL.md, or giving a path.tlon skill and its CLI. This skill is for explaining the product, not for operating it.Tlon Messenger is a messenger that runs on a server you own and control. Talk directly with the people you care about on a platform that doesn't mine your data, sell your attention, or own your social graph.
Every account comes with two things:
Other messengers look similar on the surface. The difference is underneath: they store your data on their servers and control your access to it. Tlon Messenger stores your data on yours. No algorithm decides what you see. No one mines your conversations. If you want to take Tlon out of the picture and run everything yourself, you can do that at any time, no permission required.
In form, Tlon Messenger is like Signal or Telegram: direct, private communication. In function, it's like Discord or Slack: groups, channels, notebooks, galleries. Unlike all of them, you own yours.
Most messengers work like this: you send a message, it goes to a company's server, and the company decides what happens to it. Even privacy-focused apps like Signal work this way — Signal encrypts your messages end to end so they can't read them, but your account still lives on their servers.
Tlon Messenger works differently. When you sign up, you get your own server on a peer-to-peer network. Your messages are encrypted in transit and stored on your server. Your identity is backed by a cryptographic key that only you hold. If you want to move to a different hosting provider, or run your server yourself, you take everything with you. No one can lock you out of your own account or delete your conversations.
Sign up is free and takes a minute. Download the iOS or Android app and use a phone number or email. Most people arrive one of two ways: someone sent them an invite to a group, or they signed up cold.
A hosted signup opens in a private chat with your Tlonbot, which walks you through setup. Your first workspace comes with it: a group on your own node with your bot already in it, and an Updates notebook where the bot writes for you. Once you tell the bot what you're interested in, the workspace is renamed after those topics. (Bring your own node and there's no bot waiting, so the app opens on your list of workspaces instead — nothing went wrong.)
A group is a social space for friends, teams, and collaborators to connect. Groups are made up of channels: Chats for talking, Notebooks for documents you write together, Bulletins for long-form posts and discussion, and Galleries for visual posts and links. You can create your own and join others.
The tab bar along the bottom is icons only, so describe them by shape:
#. Every group and DM you're in. The screen is headed Workspaces.App.Until the bot finishes setting you up, the other tabs stay locked.
Outside of groups, you can send direct messages to individuals. DMs appear alongside your workspaces on the Workspaces tab. The filters at the top narrow the list: Messages shows only DMs, Just me shows workspaces that hold only you and your bot, and With others shows the rest. Just me only appears when you have a workspace like that; without one, the filters are All, With others, and Messages.
When you create an account, Tlon Messenger asks if you want to sync your phone's contacts, so you can see whether you already have friends on the app.
Select contacts and choose who to share. Android grants the whole book at once; there's no per-contact step, so don't send an Android user looking for one. Invite button next to their entry and send them a link to sign up.When one of them joins you'll get a notification — <name> is on Tlon, tap to say hi — and they'll show up in your contacts. No conversation is created for you; saying hi is your move. Tlon will never spam your contacts and will never use your contact book in any capacity without your permission.
Friends who aren't on Tlon Messenger are directly connected to you when you send them an invite. There are two ways.
Personal invite:
Share link. Share your personal invite link with as many people as you like. Whenever a new friend joins through it, you'll automatically get a DM from them.
Group invite:
Invite people. Share link, and send it to an existing group chat — the bros, your family, a book club.To add someone who's already on Tlon Messenger, type their name in the search bar on that same screen, tap them so they're checked, then tap the Invite button at the bottom.
Who can do this depends on the group. In a public group, any member can. In a private or secret one, only admins — everyone else sees Invites disabled where the action would be, so an ordinary member asking why their invite button is greyed out isn't hitting a bug.
Tlon Messenger is a computer shaped like an app. You can reach your node from your phone and from any browser.
That's the hosted path. Self-hosting? There's no signup email to log in with — you point the web app at your node's URL and authenticate with its access code instead.
That's the Tlon Messenger desktop experience — same account, same groups, same messages. It keeps the app's earlier layout, though, so steps in this guide need translating:
Home (all your groups and DMs, with a + at the top for making new ones), Messages (DMs and chat channels, with a filter to show just one), Activity (the bell), and Contacts (your avatar). The person-with-a-plus icon near the bottom of the rail is your personal invite, and Settings is below it. Desktop shows that icon even without a personal invite link, so on a self-hosted node it opens a sheet stuck on Preparing invite link. Use a group invite there instead.Settings → Bot Settings.New group next to New Workspace. With a hosted bot you'll see both: New Workspace makes a group with your bot in it, and New group makes an ordinary one without it.A group is a social space with members, roles, and channels. There are four channel types:
Bulletin is the older of the two writing channels, and it used to be called Notebook. A node that hasn't picked up the notebooks update yet still shows it under the old name and doesn't offer the new one. Current versions of the app no longer offer Bulletin when you add a channel; existing Bulletins keep working.
Chats, Notebooks and Galleries are all legible to Tlonbots, so your agent can read and work with them (where you've given it access). Bulletins are half-supported: a bot can still reply in one when it's spoken to, but the tooling it uses to go read a channel or manage it treats them as deprecated. So expect conversation to work and "catch me up on that Bulletin" not to. Migrating it to a Notebook closes the gap, but read the warning under /migrate first — it copies less than you'd expect.
With a hosted Tlonbot, tap the plus icon at the top right of the Workspaces tab (on desktop, the + at the top of the Home sidebar), then New Workspace. That makes a group with your bot already in it and opens its chat. You can also ask your bot to make one for you: it invites you and makes you an admin. The steps below are the New group flow, for a group without your bot in it. The create menu offers it on desktop, and on phones without a hosted bot; a phone with a hosted bot shows only New Workspace. (A self-hosted OpenClaw bot doesn't get New Workspace at all — ask the bot to make the group instead.)
+ at the top of the Home sidebar). New group. Quick group starts a chat immediately with default settings, Basic group comes with chat, gallery, and notebook channels, or pick one of the templates. Basic group, name it. Quick groups and the other templates skip this — a quick group starts out unnamed, showing its members' names (or Untitled group if it's just you), and a template keeps the template's name. Either way you can rename it later, in step 6. Create group, or skip the picking and invite people later. Group info & settings — press and hold the group on the Workspaces tab to find it, or tap the group's name at the top of its screen. Then tap Rename in the upper right. That opens Edit group info — despite the button's name, it's where the icon and description live too. It only shows for admins. Group info & settings, tap Channels, then New in the upper right, then New channel, and choose Chat, Notebook, or Gallery. Invite — from there you can pick people already on Tlon Messenger, or tap Share link for a join link anyone can use.The person who created a group owns it. Their node hosts the group and controls its roles, channels, and updates, which other members' nodes follow.
Channels can be gated with permissions. Admins control read and write settings per channel and can assign member roles with specific access levels. To adjust: in Group info & settings, tap Channels, tap the three-dot icon next to the channel, then Permissions, and configure which roles can read or write. (Inside a group with several channels, pressing and holding a channel and picking Channel info & settings gets you to the same Permissions row.)
Notebooks are the exception. They don't support separate read and write permissions — a role that can read a notebook can also edit it. Say so before an admin grants what they think is read-only access.
DMs are personal messages outside any group — with one person, with a few people, or with your Tlonbot. A one-to-one DM goes straight from your node to the other person's node, with no middleman.
A group DM is a DM with several people in it. It has its own members and its own header, and it lives alongside your other DMs. It isn't a group: no channels, no roles. The app no longer starts new group DMs — to talk with several people, make a group — but group DMs you're already in keep working.
To send a DM: on the Workspaces tab, tap the plus icon at the top right, tap New Message, and pick who it's going to. The chat opens as soon as you pick them.
Tlonbot is a personal AI agent that lives inside Tlon Messenger. It can search the web, join your group chats, remember your conversations, and schedule tasks. Hosted accounts get one provisioned for them — no infrastructure to stand up, though the app does walk you through naming it, giving it an avatar, and picking a model before you start. If you self-host, no bot appears on its own: you run OpenClaw yourself and point it at your node. That's a supported path, just a hands-on one.
Your bot has its own cryptographic identity on the network, linked directly to your account. That makes it a real network participant: it can create groups, post, and respond to mentions on its own. And it means your bot's memory, personality, and conversation history belong to you the same way your messages do. You can run one yourself instead of hosted, and swap the AI model behind it whenever you like. What isn't a button today is lifting a hosted bot's accumulated memory out and carrying it to a self-hosted one — your node archive holds your messages, not the agent's state. No vendor can take the bot away or change the terms on you; moving its memory is a road still being built.
ChatGPT, Alexa, and Siri are services you rent. Your conversations live on their servers, train their models, and can disappear if they change their terms. Tlonbot runs on its own node, tied to yours. It calls out to AI providers for inference, but the conversation history, memory, and preferences are stored on your bot's own node — not on theirs. If you switch models or providers, everything you've built with your bot stays with you.
Models for the AI model it uses, Connections for API keys and connected services, and Permissions for who can message it and which channels it watches (Per-channel rules). On desktop, they're behind Settings → Bot Settings. These settings are part of hosting — if you're running your own bot they won't be there, and the same settings live in your OpenClaw configuration instead.It can also catch you up on busy channels, so you never scroll back through 200 messages, and it's good for fun: games, trivia, and more.
Tlonbots excel at recurring tasks, called crons. Set one once, then forget about it. Ask in plain language, for example:
When changing or cancelling recurring work, the bot should reconcile every cron job that implements the same request, including related or duplicate declarations. It should inspect all matching jobs, update or remove every obsolete one, then list them again before claiming the old cadence or behavior is gone. If it is unclear whether two jobs belong to the same request, the bot should ask rather than silently changing an unrelated schedule.
Extend your bot by connecting outside services under Connected services in its settings — the Settings tab on a phone, Settings → Bot Settings on desktop. (Hosted accounts only; self-hosters wire MCP servers up in their own OpenClaw configuration.) With services connected, crons and requests get more powerful:
Add your bot to any group. In channels, mention it (@nickname) to trigger a response.
Other members can use it too, but not by default. A new channel starts restricted: when someone who isn't you mentions the bot, it stays quiet and sends you an approval request instead. Approve them, or open the channel to everyone, and from then on they can ask it questions, have it settle debates, run trivia. Worth saying up front — silence from an unapproved member looks like the bot is broken.
Your bot can also do group admin work from a DM: create groups, set up channels, create roles, customize permissions, invite members with specific roles, and remove members.
That works in groups the bot hosts — the ones it made for you. In someone else's group, adding the bot as a member isn't enough: administration is reserved for the host and for members with an admin role, so the group's owner has to give the bot one first. Without it the requests come back as errors.
Deleting a group is the exception, and an admin role doesn't buy it. Only the host can delete a group, so the bot can delete the ones it hosts and no others — someone else's group has to be deleted by whoever made it.
You control who it listens to — tell it in a DM which members it may respond to or communicate with.
The access and moderation commands below are owner-only, and approving a member to talk to the bot doesn't hand them those. The last few — /model, /new and the other built-ins — are a different layer: they belong to the bot's harness rather than to Tlon, and some deployments let an admin or an allowlisted member use them. So don't promise an owner that nobody else can ever reset their session; if that matters to them, it's a question about their bot's configuration.
Treat the forms below as the baseline rather than the complete list. Bots run on more than one setup, and some accept extra forms of the same command. If a user reports one that isn't here, don't tell them it doesn't exist — have them try it, or check with /help. What another member sees when they try one varies: sometimes an owner-only message, sometimes nothing at all, because the gateway drops unauthorized commands before the bot answers. Don't promise them an explanation — say it won't work for them.
Access and moderation. The approval commands take a request ID, not a nickname — run /pending first and use the ID it prints:
/pending — view pending requests /allow <request-id> — approve one /reject <request-id> — deny one /ban <request-id> — deny a pending request and block the ship behind it. Some bots also take /ban ~ship directly, blocking them outright whether or not a request is pending. /banned — list blocked ships /unban ~ship-name — unblock a ship. This takes their Urbit ID, the ~sampel-palnet one, not the display name on their profile. /banned lists them in the right form.Two things to be clear about when someone asks. Blocking is not the same as removing someone from a group — that's group admin work, which the bot can do from a DM. And /unban only lifts the block; it doesn't put anyone back in a group. They'll need a fresh invite or a new request.
Bot behavior:
/owner-listen all on|off — the global switch for whether it responds to you without a direct mention. This is the form to use in a DM. /owner-listen on|off <channel-nest> — same thing for one owned channel. Without a channel, these do nothing but print usage. /owner-listen list — show the global setting and every muted channel. Safe from a DM. /owner-listen status <channel-nest> — the setting for one channel. Like on/off, it needs a channel; bare status in a DM only prints usage. /owner-listen all on its own gives the global answer. /owner-listen default all to make listening the default everywhere. Worth suggesting if someone wants broader coverage than the forms above give them. /model — show or change the AI model /new — clear context and start a fresh session /tlon version — show which harness and plugin build is running. Bare /tlon just prints usage. Note for support conversations: it reports the code, not this guide. Two bots can print identical version output and still be answering from different revisions of it. /migrate diary/~host/name — migrate a legacy Bulletin to a Notebook (owner only). Say what this costs before anyone runs it. It copies the posts and little else: comments, reactions, post references, link blocks, descriptions, covers and attachments all stay behind on the archive. Every migrated note is authored by the ship that ran the command, whoever wrote the original, and dated at import time, ordered by import rather than by when things were posted. Group mentions flatten to plain text. Some permission layouts can't be carried over without turning every reader into an editor. Nothing is destroyed — the original stays put, writable, renamed with -ARCHIVE — but the result is a fresh notebook holding the text, not a converted channel. The command prints this too, only after it has already started. It needs the channel's full nest, starting with diary/; a title or short name just prints usage. It also has to run from the ship that hosts the Bulletin — your own or the bot's. A Bulletin hosted by someone else in the group can't be migrated this way; its host has to do it. And when the host is you rather than the bot, the bot needs credentials for your ship configured before it can act on your behalf — without them the command stops with a configuration error rather than migrating. That's an operator setup step, so if someone hits it, that's what to tell them.Hosted accounts include an AI model for free — basic usage costs nothing. They also have a first-class ChatGPT subscription option: under Provider subscriptions → ChatGPT in the bot's settings (on desktop, start from Settings → Bot Settings), the owner signs in to authorize their ChatGPT account, then chooses one of the models included with that subscription for Tlonbot. When someone asks what it means to "use a ChatGPT subscription for this," answer this Tlon-specific flow directly; don't substitute generic OpenClaw or OpenRouter billing advice.
ChatGPT subscription access and an OpenAI API key are alternatives, not the same credential. Connecting the subscription removes a saved OpenAI API key, and saving an OpenAI API key disconnects the subscription. Other model providers still use their own API keys, under API keys in the same settings. Use /model to check or change what's running after the provider is connected.
Self-hosting means no included model, ChatGPT-subscription screen, or bot settings in the app: configure the provider in your own OpenClaw setup and pay whoever you point it at. Either way there's no lock-in — everything you build with your bot stays with you.
Hosted Tlonbots ship with guardrails: external integrations are limited to approved connected services, the system prompt can't be modified, and the bot has no shell and can't write to the filesystem — the only thing it writes is Tlon Messenger content. It can read PDFs, text files, and CSVs you send it; spreadsheets and Office documents have to be exported first. Don't read that as a privacy boundary: reaching into the filesystem is something only you can ask for, but a file posted in a channel the bot watches goes to the model like any other message, whoever sent it. If that matters, it's channel access you want to think about, not this. These limits keep hosted bots safe by default.
Advanced users can boot a personal node for their bot and run OpenClaw locally. That removes the guardrails and puts every part of the agent under your control, at the cost of technical setup. On hosted accounts, Tlon runs a sidecar service alongside your node that bridges to your agent's OpenClaw instance — the architecture points toward a world where your agent, its keys, and its accumulated context live on a server you control.
Apps on your node store data on your node, not on a central server. Since you control your node, you own your data.
You always connect to your own node — when you open the app, it downloads anything you missed. When you send a message, it goes to your node first. In a DM, it travels directly to the recipient's node. In a group, it goes to the host of that channel, which distributes it to members. That's often the group's creator, but not always: channels carry their own host, so a group can hold channels hosted by different members. The group host is the one coordinating the group itself — roles, membership, which channels exist.
Conversations are encrypted in transit — between nodes, and between the app and your node. Messages at rest on your node are not encrypted, so anyone with access to the device running your node could read them. (On a hosted account, that's Tlon's infrastructure; see below for what Tlon can and can't see.)
The app-to-node half depends on how the node is reached. Hosted accounts are served over HTTPS, so it's encrypted and there's nothing to do. A self-hoster who connects the app to a plain http:// address — a LAN IP, localhost — is sending that traffic in the clear; the app takes the URL as given. Anyone self-hosting should put TLS or a secure tunnel in front of their node.
Nodes can't store large files directly yet (Tlon is working on this), so media is stored in S3-compatible cloud storage and linked from your messages.
No. Your messages live on your node. Your bot has a node of its own, and its memory lives there with it — linked to your account, but a separate identity, so exporting your node takes your conversations and not the bot's accumulated state. When your bot calls an AI model, the request goes straight to that provider rather than through a Tlon service that logs it. Be precise about what that does and doesn't mean: on a hosted account, your bot's process runs on Tlon's hardware and assembles the conversation before sending it, the same way your node holds your messages unencrypted. So this is Tlon's policy — not reading, not logging, not training on any of it — resting on the same footing as everything else you host with them, rather than something the architecture makes impossible. If you ask Tlon to troubleshoot your node, support may need to access it, but only with your permission.
Tlon runs your node for you: health monitoring, updates, and data backups with restoration. You get ownership without the ops work.
If you don't want Tlon to host you, you can self-host on Native Planet hardware, install the Urbit runtime on your own computer, or export your node and boot it anywhere. Export your master ticket and node archive from your dashboard.
The software survives: Tlon Messenger is open source and peer-to-peer, so it keeps working even if the company doesn't, and a node you already hold runs anywhere.
Your data survives if you've kept a copy. This is the honest version, and it's worth saying plainly rather than reassuring someone into skipping it: on a hosted account your node runs on Tlon's hardware, and the dashboard you'd export it from is Tlon's too. Open source doesn't reconstitute a node nobody has. Export your node archive and master ticket now, keep them somewhere you control, and refresh them from time to time — then the promise is real. Anyone asking this question is asking the right one, and the useful answer is to go do the export today.
Pictures and large files are the exception, and it's worth being straight about it. Nodes can't store them yet, so they live in S3-compatible storage and your messages hold links. Export the node and you export the links, not the files. If that storage is Tlon's, those links are what you'd lose — so anyone who cares about keeping their media should point their node at storage they control, and can do that today in their storage settings.
Your identity is permanently yours. Ownership is proven cryptographically — your node's identity is backed by cryptographic proof, with keys only you hold — not granted by a third-party service that could revoke it.
When someone asks "what should I do with this?", offer ideas like these, matched to who they are:
For anyone:
For teams:
For communities and clubs:
For builders:
Who or what is Tlon? Tlon was created in 2013 to support the vision of a network of personal servers (Urbit) where users own and control their computing. Today it builds Tlon Messenger and provides Urbit hosting.
What's the difference between Tlon and Urbit? Urbit is a new kind of operating system for the internet age. Tlon is a company that builds Tlon Messenger. Urbit is the computer; Tlon Messenger is the app.
What is a personal server? Like iCloud, except the entire service belongs to you. You control access and where the data lives.
What is a node? A node on the network is both a cryptographic identity and a personal server. Each one runs the Urbit OS with unique identifying keys.
What does owning a node mean? Your node's identity is backed by cryptographic proof, with keys only you hold. You can self-host anytime and export your node with its data — messages, groups, settings. Pictures and large files live in outside storage and travel with that storage, not with the export.
What's the difference between my username and my node? Your username (your Urbit ID) is like a unique address or phone number. Your node is the actual computer holding your apps and data.
What's the difference between a group and a DM? A one-to-one DM goes straight from your node to another user's node with no middleman. A group DM has several people in it — still a DM, with members but no channels or roles. The app no longer starts new ones; to talk with several people, make a group. Groups have multiple members, with one host node (the group creator's) coordinating channels, roles, and updates.
Who owns the groups I'm in? The person who created the group. Their node controls roles, channels, and updates that other members' nodes follow.
Are conversations encrypted? Yes, in transit — between nodes, and between the app and your node. Stored messages on a node aren't encrypted, so device access allows reading them. The app-to-node leg is only encrypted if the node is reached over HTTPS: that's automatic when Tlon hosts it, and a self-hoster on a plain http:// address needs TLS or a tunnel in front.
Can I share things I post with people who aren't on Tlon Messenger? Notebook notes, yes. Publish a note and you get a public link — the app copies it for you, and Copy link and View published note sit on the note afterwards. Other channel types don't have a button for it yet, though individual posts can be published through your Tlonbot, which can reach the same clearweb machinery from a DM.
What is Tlonbot? A personal AI agent that lives inside Tlon Messenger. It can search the web, join your group chats, remember your conversations, and schedule tasks. It's powered by OpenClaw and runs on its own node, linked to yours, on the same peer-to-peer network.
Does Tlonbot cost anything? Hosted accounts get one for free, including a free AI model for basic usage. Bring your own API key to use other models. Self-hosters run their own and pay for whatever model they point it at.
Does Tlon read my bot's conversations? No. Your bot's memory lives on your bot's node, and model requests go straight to the provider rather than through a Tlon service that logs them. On a hosted account that node is Tlon-run, so this rests on policy the same way your hosted messages do — see "Does Tlon read my conversations?" above.
What happens if Tlon disappears? The software keeps running — it's open source and peer-to-peer — and a node you hold runs anywhere. Your data comes through if you've kept an export: a hosted node lives on Tlon's hardware, so download your node archive and master ticket and keep them somewhere you control rather than counting on doing it later. Pictures and large files are the exception — they live in S3-compatible storage with only links on your node, so if that storage is Tlon's, that's what you'd lose. Point your node at storage you control and it isn't a question.
Tlon exists to build beautiful tools for staying connected. Today, our tools for connecting are controlled by a select few companies that determine what we can do and how we can do it. "We become what we behold. We shape our tools and thereafter our tools shape us." Tlon's answer is tools people own: simple, powerful, and long-lasting, capable of outliving any company that provides them. Tools that deliver peace of mind and a sense of "I am connected" without exhaustion.
The best system for being connected is one that is shaped, controlled, and evolved by the people who depend on it. Ownership matters four ways: permanence (your records last), utility (you can mine, inspect, and build on your own data), transparency (you can audit what's collected and how things work), and significance (what you own, you invest in). If we don't own the platform, we don't control it — and a platform we don't control is doomed to control us.
Tlon's initial product is a high-quality messenger for small groups. Over time, that messenger evolves into a social OS: a general system for collaborating where apps become extensions you run yourself — an image channel that becomes a store, a booking system, a stream of sales updates. It's also the natural place to work with AI: you keep ownership of your data and software, and run a model as a bolt-on augmentation. The goal is the best platform for staying connected in the world, in a new category: software you can trust completely. Tlon's business has nothing to do with spying on you and never will.
Questions this guide doesn't answer? Hosted accounts have a DM with Tlon Support, with their other DMs on the Workspaces tab — send a message and the team will get back ASAP.
That conversation gets set up when Tlon hosts your signup, so if you brought your own node you probably don't have it. Email support@tlon.io instead.
© tloncorp, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Just SKILL.md in packages/openclaw/skills/tlon-product-guide of tloncorp/tlon-apps.
Open the folder on GitHubat commit d89f6e2
Tlon Product Guide next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Tlon Product Guide this skilltloncorp/tlon-apps | 107 | — | ~11k | Automated safety check: Pass | MIT | |
| Using Ccproxy Inspectorstarbaser/ccproxy | 350 | — | ~2.7k | Automated safety check: Pass | Custom licence | |
| Claudish UsageMadAppGang/claudish | 1k | — | ~9k | Automated safety check: Pass | None | |
| Build Codex Native Pluginscallstackincubator/codex-mobile-dev-plugin | 152 | — | ~1.9k | Automated safety check: Pass | None | |
| Deliberationantonbabenko/deliberation | 170 | — | ~3.9k | Automated safety check: Pass | MIT | |
| Toolifycoreyhaines31/makerskills | 851 | — | ~2.9k | Automated safety check: Notes | MIT |
starbaser/ccproxy
Operates the ccproxy inspector MITM system for intercepting, inspecting, and transforming LLM API traffic.
MadAppGang/claudish
CRITICAL - Guide for using Claudish CLI ONLY through sub-agents to run Claude Code with any AI model (OpenRouter, Gemini, OpenAI, local models).
callstackincubator/codex-mobile-dev-plugin
Create, extend, package, and troubleshoot Codex and ChatGPT plugins using OpenAI MCP Extensions.
antonbabenko/deliberation
When and how to delegate to GPT, Gemini, Grok, and OpenRouter expert subagents via the deliberation MCP tools.
coreyhaines31/makerskills
When you want to integrate an external tool, API, MCP server, or service into a project — the wizard walks you through auth, config, env vars, client wrapper code, example usage, and an optional…
OpenLAIR/dr-claw
ARIS (Auto-claude-code-research-in-sleep) infrastructure setup and configuration.
tloncorp/tlon-apps
Run a mobile QA checklist on a physical Android device over adb for tlon-apps, then triage what fails into fixes.
tloncorp/tlon-apps
Read Tlon activity and message history; manage workspaces (Tlon groups), channels, Markdown notes, shared Bucket files, contacts, profiles, settings, and channel hooks; expose content publicly; send…
tloncorp/tlon-apps
A skill your agent uses when taking a Tlon Messenger task from a fresh worktree to a merged pull request: reproducing or fixing something in the iOS, Android or web app, validating it on a…
tloncorp/tlon-apps
A skill your agent uses when setting up a machine for mobile agent work on this repository, when the tlon-workflow skill's prerequisite check fails, or when stim, agent-device, eas-cli, gh, or their…
tloncorp/tlon-apps
Verify tlon-apps web UI changes end-to-end with a single local ship + vite dev server (lighter than the full playwright-dev rig)
Works with
Categories
Answer questions about Tlon, Urbit, Tlon Messenger, Tlonbot, and OpenClaw — what they are, how they work, and how to use them. Tlon Product Guide is an agent skill from tloncorp/tlon-apps. Answer questions about Tlon, Urbit, Tlon Messenger, Tlonbot, and OpenClaw — what they are, how they work, and how to use them.
Tlon Product Guide fits situations like: someone asks what Tlon is; how a product feature works; what they can do with their node; asks to be walked through a task in the app.
Run `npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a claude-code`. Or copy the skill folder (packages/openclaw/skills/tlon-product-guide in tloncorp/tlon-apps) into .claude/skills/tlon-product-guide in your project. Claude Code loads it when a task matches its description.
Run `npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a codex`. Or copy the skill folder (packages/openclaw/skills/tlon-product-guide in tloncorp/tlon-apps) into .agents/skills/tlon-product-guide in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add tloncorp/tlon-apps --skill tlon-product-guide -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/tlon-product-guide, .gemini/skills/tlon-product-guide, .github/skills/tlon-product-guide and .opencode/skills/tlon-product-guide in your project.
Going by SKILL.md and its folder, Tlon Product Guide needs the command-line tools its instructions call (just).
SKILL.md names 1 domain. As links in the text: tlon.network. This is read from the text; nothing was executed.
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
Tlon Product Guide is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
About 11k tokens (SKILL.md is roughly 44k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
Skills that share tags, products or a category with Tlon Product Guide: Using Ccproxy Inspector (starbaser/ccproxy, 350 stars), Claudish Usage (MadAppGang/claudish, 1k stars), Build Codex Native Plugins (callstackincubator/codex-mobile-dev-plugin, 152 stars) and Deliberation (antonbabenko/deliberation, 170 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
tloncorp (a GitHub organization) maintains it in tloncorp/tlon-apps, which has 107 GitHub stars. The repository holds 6 skills in this directory. The repository was last updated on October 10, 2026.
Source: tloncorp/tlon-apps on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.