Every AI your team uses,on the same page.
0bridge is one MCP endpoint that gives every AI tool your team uses, from Claude and ChatGPT to Claude Code, Codex and Cursor, the same services, files, instructions and skills. Choose your team's services once; everyone signs in with their own account and keeps their own memory and history. Free for one person.
One MCP serverfor every AI tool you use
Sign in to a service once. 0bridge keeps the login in the cloud and calls it for you, so every tool sees the same services, on every machine and every account.
One endpoint
Linear, Notion, Sentry or any MCP URL, through one server. Apps list four tools that find and run the rest, so connecting more never crowds their list.
Every machine
Sign in on a new laptop and your connections are already there. No config to copy.
Every account
Work and personal side by side. A linked repo picks its own account.
Connect
Sign in to each service once. Logins stay sealed on 0bridge; you can turn single tools off or make a connection read-only.
Apply a skill
Your skills and a work folder's instructions, the same in every tool you connect. Skills
Run what's allowed
Reads and changes run through separate tools, so apps ask before a change. Tools you turned off are refused.
Keep the result
Results go back to Drive, in the work folder every tool reads.
Services that send webhooks connect here too: give Channel Talk, GitHub or any service an address, and each event runs a command on your machine or wakes a session that listens for it. Webhooks
Paste one promptYour agent does the rest
It installs the CLI and wires up Claude Code, Codex, Cursor and Gemini CLI on your machine. You only approve sign-ins in the browser.
Connect a service once
Every tool gets its tools the moment you sign in.
Let your agent set it up
Paste this into Claude Code, Codex, Cursor or Gemini CLI.
Set up 0bridge on this machine for me. 0bridge gives all my AI tools (Claude Code, Codex, Cursor, Gemini CLI) one MCP endpoint, https://0bridge.dev/mcp, and keeps my service sign-ins (Linear, Notion, …) in the cloud, so I sign in once for every tool. 1. Install the CLI: `npm install -g 0bridge@latest` (needs Node 20.12+). Check it with `0b --version`. 2. Run `0b setup --yes`. It finds the AI tools on this machine, signs me in, and adds 0bridge to each of them, backing up every file it changes. - Sign-in opens my browser. Show me the one-time code from the output right away, then wait: the command finishes by itself once I approve (up to 10 minutes). Give it a long timeout, or run it in the background and watch its output. 3. Ask me which services to connect. `0b connect` lists the known ones (linear, notion, sentry, github, context7, …); any MCP URL works too. For each one, run `0b connect <service>`: it opens that service's sign-in in my browser and waits until I finish. For a second account of the same service, add `--label <name>`, for example the workspace name. 4. Run `0b tree` and show me the result. 5. Tell me to restart my AI tools, this session included, so they load 0bridge. Only use the 0b CLI for this; don't edit tool config files by hand.
Or run it yourself
npm install -g 0bridge@latest
0b setupClaude and ChatGPT
Add a custom connector with this URL, then sign in and allow access.
https://0bridge.dev/mcp
Sign in
0b setup opens your browser with a one-time code. Approve it; this machine gets its own token.
Every AI tool gets 0bridge
Claude Code, Codex, Cursor and Gemini CLI on this machine get one MCP server and the 0bridge skill, so later requests like "add the Sentry MCP" go through 0bridge too. Config files are backed up first.
Connect services once
0b connect linear opens Linear's sign-in. From then on every tool has Linear's tools, as linear__….
The tools you already useand how well each one works
Five levels, each dated. Where we haven't tried a product, we say so instead of guessing. The full list
Full
- Claude Code
- Codex
- Cursor
- Gemini CLI
0b setup adds 0bridge to them. Their history is collected, you can resume sessions into and out of them, and Claude Code and Codex can be driven from your phone.
Connector
- Claude (web, desktop, mobile)
- ChatGPT
- OpenAI Dots
Add a custom connector and sign in. Your services, profile and memory are there; your coding history too, which new accounts start with on and you can turn off.
Checked 2026-10-01 · Dots 2026-10-05Bearer token
- Grok and Grok Bot
- Any MCP client
A URL and a scoped token from the dashboard. Tools only by default: no history, no agent control, 90 days.
Checked 2026-10-01Untested
- Meta Muse and Muse Code
- Manus
- Perplexity
- The Gemini app
We haven't tried these on the real product, so we don't say they work. Tell us if you do.
Checked 2026-10-01Not supported
- Instinct
No terminal and no MCP yet, as far as we can tell, so there is nothing to connect. Not supported yet.
Checked 2026-10-01Resume anywherein any tool, on any machine
Copy one short reference and the next tool picks up the goal, the open items and the last turns. Nothing is stuck in the app you started in. How resuming works
One short reference
The copy icon on a session in the dashboard copies "continue 0b:k3f9x2".
- Any connected AI can resolve itA coding agent calls
bridge__resume. A chat app can too while chat history is on: new accounts start with it on, and you can turn it off. - The same handoff every timeBuilt without an AI model on our side: your first message, any saved summary, and the last turns.
- Checkpoints from any toolAsk a tool to save a summary and the open items before you switch.
- The honest limitThe next tool starts a new session that knows what the old one did. It doesn't move a live process.
Your agents from your phonewithout a bot to host
Start and steer Claude Code and Codex on your own machine from Claude, ChatGPT or the dashboard, and approve their permission prompts. It replaces a Slack bot or a gateway for this one job. How it works
Outbound only
A small program on your machine connects out to 0bridge. No open port, no tunnel, nothing to host.
Nothing runs until you say so
New accounts start with the switch on, but a machine runs nothing until you turn it on there, and only in repos you list. Each task gets its own git worktree. Pushing to main, merging and deploying are denied by default.
Approvals in your hand
Permission prompts are relayed to you, not answered for you. A notification links to the task page; there is no approve button in a notification.
Which session needs youon every machine, in one list
Your coding agents' hooks tell 0bridge when a session is working, waiting for you or done. See it on the dashboard's Now page, with 0b sessions, or ask any connected AI. How the board works
An example of the Now list, not live data.
- Claude Code, Codex and CursorClaude Code reports when it waits for a permission or an answer (not yet seen on a live session, 2026-10-01). Cursor has no permission hook, so it never shows as needing you. Codex reports what its hooks allow.
- Up to two short lines, redactedText shaped like a key or token is masked first. With end-to-end history, no lines or titles are kept, and your machines stop sending them.
- A nudge only if you askA notification when a session starts waiting is off until you turn it on, and comes at most once per session every 2 minutes.
- Not live to the wordThe board changes when a hook fires, a few seconds apart. There is no per-token streaming.
What your agents usetokens, never text
Token counts per tool, model, repo and day, with an estimated cost. What is collected
Counts only
Input, output and cache tokens from the logs your tools already write. No prompts and no answers. The server can read the counts, even with end-to-end history.
An estimate, labeled
Priced at published API list prices, with their date shown. Subscriptions don't bill per token, and a model with no price shows a dash.
Yours to turn off
0b usage off stops uploads, and Forget usage on the dashboard deletes what is stored.
Your context follows youacross tools, accounts and machines
Past conversations, and a profile, memory and skills every AI can read. Profile, memory and skills
Search every conversation beta
Claude Code, Codex, Cursor, Gemini CLI and Grok sessions in one place, once you turn it on. Your coding agents can search them, and you can continue any of them in another tool.
- One profile and memoryEvery AI you connect reads the same profile and memory, and saves to it when you ask.
- Your skills and instructionsIn chat apps over MCP, and in Claude Code, Codex, Cursor and Gemini CLI with
0b apply.
We can't read what you sealand we say what we can
End-to-end history is encrypted on your devices with a key only your devices hold. What the server can read, it reads for a reason: here's exactly what and why.
Sealed on your devices
Conversation history in end-to-end mode is encrypted before it leaves your machine. We can't open it.
Logins used only for your calls
Service logins are encrypted at rest with a key per user, and used only to make the calls your tools ask for.
Calls logged, never their content
The activity log keeps each tool call's name, result and time, never what was sent or what came back. Rows are deleted after 90 days.
Source-available client
The 0b CLI, and the code that encrypts your history, is on GitHub for anyone to read. See exactly what runs on your machine.
Start freePlus $6 a month, Team $15 per seat
Every connection on every plan. Free has 5,000 tool calls a month; Plus has no monthly cap; Team is $15 per seat after a 30-day free trial.
Free
Every connection, 5,000 tool calls a month.
- One MCP server for Claude Code, Codex, Cursor, Gemini CLI, Claude and ChatGPT
- Any service, and several accounts of each
- Conversation history, and resuming a session in any tool
- Profile, memory and skills in every tool
- Your coding agents from your phone, on the machines you turn it on for
- A board of what every session is doing, webhooks that wake a session or run a command, and token usage
Teamearly access
Team connectors, shared skills and a team drive are live; the admin view is being built with our first teams.
- Everything in Plus, for each member
- Team connectors: everyone signs in with their own account
- Shared skills and instructions in every member's AI tools
- A team drive your agents search and read, with generous storage
- Members, audit log and usage in one place (being built)
Plus is $6 a month ($60 a year) with no monthly cap on tool calls. A seat is a member of a team workspace; teams start with 30 days free. Enterprise is a contract: a SOC 2 report for 0bridge and a HIPAA BAA come with it, and SSO is planned. Compare plans
Is my context locked into 0bridge?
No, and that's the point. Your history stays in the files your tools wrote, your instructions and skills are Markdown files on your machine, and 0bridge keeps an index you can delete. The session format is open (0b.session/1, Apache 2.0). Delete your account and what's on your machines stays. Why no lock-in
Is it free?
There's a free plan with every connection and 5,000 tool calls a month, no card and no time limit. Plus is $6 a month with no monthly cap, and Team is $15 per seat after a 30-day free trial. Pricing
What can the server see?
The logins of the services you connect (stored encrypted, used only to call them for you), the calls passing through (the log keeps each tool's name, result and time, not the content), your conversation history if you turn it on, your profile, memory and skills (readable, because chat apps have to read them), and agent task text if you use agent control. It can't read your history in end-to-end mode: it's encrypted on your machine with a key only your devices hold. Security
Which AI tools and services work?
0b setup adds 0bridge to Claude Code, Codex, Cursor and Gemini CLI on your machine. Claude and ChatGPT add it as a custom connector (ChatGPT needs a paid plan with developer mode), and OpenAI Dots uses ChatGPT's. Grok and any client that takes a URL and a token use an app token. We haven't tried Meta Muse, Manus, Perplexity or the Gemini app, and the client list says so, with the date. Instinct has no terminal or MCP yet, so it is not supported. For services: about 60 are known by name (Linear, Notion, Slack, GitHub, Sentry, Stripe, …), and anything else with an MCP server, an OpenAPI document or an API key works too. Supported clients
Can Claude or ChatGPT see my coding history?
While chat history is on. New accounts start with it on; any app can turn it off when you ask, and so can the dashboard. Turning it back on is only from the dashboard after you confirm it's you, and it is never offered through a connector directory listing. Coding agents like Claude Code and Codex can search it once you've turned history on. Conversation history
Can I start coding agents on my machine from my phone?
Yes, on a machine where you turn it on and list the repos it may use (new accounts start with the account's switch on; each machine stays off until you turn it on there). Each task runs in its own git worktree, dangerous commands are denied by default, and permission prompts come to you. Your agents from your phone
Can I see which coding session is waiting for me?
Yes, on the dashboard's Now page, with 0b sessions, or by asking a connected AI. Hooks report each session's state with up to two redacted lines of text, and none at all with end-to-end history. Cursor has no permission hook, so it never shows as needing you. Session board
Does it work on Linux, on Windows, over SSH?
The 0b CLI runs on macOS, Linux and Windows with Node 20.12 or newer. macOS and Linux are the most used; Windows support is new, so expect rough edges and tell us about them. Signing in uses a one-time code you approve in any browser, so it works over SSH. The dashboard, Claude and ChatGPT work anywhere.
What happens to my data if I leave?
Disconnecting a service deletes its tokens right away, and 0b history forget --all deletes your history. Delete your account in the dashboard under Settings: it deletes your account, your connections and their stored tokens, your vault, conversation history, personal files, profile, memory, skills, agent data and projects. Encrypted backups age out within 28 days (weekly ones within 12 weeks). 0b backs up each AI tool's config before changing it, so 0b restore puts it back.