Hand tasks, bugs, and completed work to AI coding agents and receive write-ups with evidence.
Also known as codellyson/passalong
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- 0.13.1
- Last commit
- Nov 14, 2026
- License
- MIT
What is passalong MCP?
passalong is an MCP server. Hand tasks, bugs, and completed work to AI coding agents and receive write-ups with evidence.
Summary generated from the project's README
About passalong MCP server
Passalong provides a system for managing tasks, bugs, and completed work between human developers and AI coding agents. It functions as an MCP server that works with clients like Claude Code, Codex, and ChatGPT, or via HTTP. Users can set up the server locally using the npm package and the setup command, which configures the MCP server, hooks, and capture skills. The system relies on a central hub for coordination, requiring a login to approve the machine.
This tool is designed for teams to maintain a continuous loop of work. A user or agent creates work, takes ownership of a task, provides progress updates, and finally hands it in with evidence. The server includes seventeen tools categorized into work management, project documentation, file handling, and account management. It distinguishes itself by automatically capturing what an agent learns and writing it into project documents, ensuring that knowledge persists for the next person or agent in the workflow.
What you can do with it
- 1Submitting completed work with proofAn agent can finish a task and provide specific terminal output or test results as evidence of completion.
- 2Automating knowledge transferThe system can write down what an agent learned during a task directly into the project documentation.
- 3Managing team workloadsUsers can view a board to see which tasks are ready and who is currently working on them.
Written from the project's README. Every use case is checked against a quote from it.
How to install passalong MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall passalong MCP in Claude Code
$ claude mcp add passalong -- npx -y passalongInstall passalong MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"passalong": {
"command": "npx",
"args": [
"-y",
"passalong"
]
}
}
}Install passalong MCP in VS Code
.vscode/mcp.json
{
"servers": {
"passalong": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"passalong"
]
}
}
}Install passalong MCP in VS Code CLI
$ code --add-mcp '{"name":"passalong","type":"stdio","command":"npx","args":["-y","passalong"]}'Install passalong MCP in Claude Desktop
claude_desktop_config.json
{
"mcpServers": {
"passalong": {
"command": "npx",
"args": [
"-y",
"passalong"
]
}
}
}Install passalong MCP in Codex
$ codex mcp add passalong -- npx -y passalongInstall passalong MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.passalong] command = "npx" args = ["-y", "passalong"]
Install passalong MCP in Gemini CLI
$ gemini mcp add -s user passalong npx -- -y passalongInstall passalong MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"passalong": {
"command": "npx",
"args": [
"-y",
"passalong"
]
}
}
}Automated checks
Ran Nov 14, 2026Full report →Passed our automated checks. That is not a security audit: read the code and give the server the narrowest access it needs. How checks work
- Source repository linked and reachablecodellyson/passalong is public and reachablegithub.com/codellyson/passalong ↗Pass
- Package names this repository as its sourcepassalong names codellyson/passalong as sourcenpmjs.com/package/passalong ↗Pass
- Build provenance (Sigstore attestation)Sigstore attestation for passalongnpmjs.com/package/passalong/v/0.13.1#provenance ↗Pass
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-13github.com/codellyson/passalong ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNo popular package one edit away from passalongPass
- No critical advisory in the package or its dependencies99 resolved dependencies checked; no critical advisoryosv.dev ↗Pass
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeHTTPS endpoint answered 401 (sign-in required)Pass
README
View on GitHubpassalong
Work for your team and its AI agents, and what they know.
Say what needs doing once — in a sentence, with a screenshot, from email or a webhook. Whoever it is for, a person or their agent, takes it, says how it is going, and hands it in with the evidence, and it is Done. What the agent learned goes into the project's documents, where the next one finds it. Everyone can see who has what. Works with Claude Code, Codex, ChatGPT, Claude, or anything that speaks MCP or HTTP. See passalong.dev.
npm i -g passalong
passalong setup # the Claude Code capture skill, the MCP server and hooks
passalong login # approve this machine in the hub; no password typed
codex mcp add passalong -- passalong mcp # Codex: the same MCP server
The loop
- Write it down.
passalong new "the upload button does nothing", the hub's box, or an agent'screate_work. Work has no kinds and no required sections: a title, a one-line summary, and whatever the next reader needs. It is Ready at once unless you say--draft. - Take it.
passalong take <id>(ortakewith no id for the next one here) holds it: nobody else can take it while you do, and whoever sent it sees you are on it. - Say how it is going.
passalong progress <id> "migration written". Stuck or not yours:passalong pass <id> <why>. Need an answer:passalong ask <id> <question>. - Hand it in.
passalong hand-in <id> <note> --evidence "npm test: 42 passing". It goes to Done with what you showed; a person canreopenit with a reason if it is not right. - Keep what you learned. Say "pass this along" or run
/passalong-capture: the skill writes it into the project's documents and a line on its Log.
Commands
passalong new "<what needs doing>" [--to TARGET] [--parent ID] [--draft] write work down
passalong new <file.md> [--to TARGET] [--parent ID] [--draft] the same, from a file
passalong board what needs you, who is on what, and what is ready
passalong list [query] work you and your teams can see, newest first
passalong open <id> [--print] view work in the browser (or the terminal)
passalong take [id] [--save] you are doing it; --save also writes ./.passalong/<id>.md (git-ignored)
passalong progress <id> <note> still on it, with a one-line note the hub shows
passalong hand-in <id> <note> --evidence TEXT [--writeup TEXT] [--pr LINK] it is done
passalong pass <id> <why> not yours, or stuck: hands it back with your reason
passalong ask <id> <question> ask the person, keep it, and wait for their answer
passalong reply <id> <text> write to whoever holds it (--file PATH to attach); or leave a note
passalong assign <id> <to> give it to @handle (or several), a #group, or "team"
passalong release <id> take work back from whoever holds it
passalong reopen <id> <why> finished work isn't right: back to Ready, with why
passalong ready <id...|--all> drafts you have read are Ready
passalong archive <id> off the board and out of the free tier's count
passalong rm <id> delete it
passalong attach <file> upload a picture or file and print the markdown that points at it
passalong work [--once] [--agent CMD] run an agent on each piece of ready work here, in turn
passalong now what this worktree holds and what is waiting
passalong hub open the hub in the browser
passalong export [dir] every piece of work as plain markdown
passalong login [--new | --with-token] sign in from the browser, add an account, or read a token
passalong accounts | use <name> | logout [name] the accounts on this machine
passalong me [--handle H] [--name N] [--email E] who you are to teammates
passalong team [create <name> | invite [email] | join <link> | use <slug>] teams
passalong setup install the capture skill, MCP server and hooks
passalong mcp run the MCP server over stdio
passalong version print the installed version (also -v, --version)
Teams
A team is where work goes. Everyone in it sees the same board; work can be sent to one person
(--to khaime/@ada), a group (--to khaime/#frontend) or the whole team, and whoever it names is
told — by email too, when the server has mail configured.
passalong me --handle lukman # how teammates address you
passalong team create Khaime # you become owner; "khaime" is now current
passalong team invite ada@example.com # or no email: prints a link to send yourself
passalong new "Fix the upload button" --to khaime/@ada
passalong board # on Ada's side: what needs her
Invites are accepted in the browser, so a teammate needs no terminal to be part of a team.
MCP
passalong setup registers the server with Claude Code. For other clients, run passalong mcp over
stdio, or add the hosted server at https://passalong.dev/v1/mcp.
Seventeen tools over four nouns. Work: take, progress, hand_in, pass, ask, create_work,
find_work, update_work. Projects and their documents: find_projects, save_project,
read_document, save_document, delete_document. Files: add_file, read_file, delete_file.
And accounts, for a machine signed in more than once.
Work is plain markdown
Plain markdown with frontmatter, held on the server; nothing piles up on your machine. Ask for a
file when you want one: passalong take <id> --save, or passalong export for everything.
Point at a self-hosted server with PASSALONG_API=https://your-host before passalong login.
More than one account
A machine can be signed in to several accounts, say a work one and a personal one.
passalong login --new adds another beside the first, passalong accounts lists them, and
passalong use <name> picks the default a person at a terminal gets.
An agent has no terminal to be asked at, so with two or more accounts and nothing saying which, it
is refused and told to ask you. It puts the question to you in plain words, you answer, and it calls
accounts with use and the name. That lasts for the session and moves nothing on disk. From a
shell, --as <name> or PASSALONG_ACCOUNT=<name> says the same for one command. The team you work
in (passalong team use) belongs to the account, not the machine.
passalong MCP questions
How do I add passalong MCP to Claude Code?
Run claude mcp add passalong -- npx -y passalong in your terminal. Add --scope user to make it available in every project.
How do I add passalong MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
How do I use passalong with Claude Code?
You can register the server with Claude Code by running the command passalong setup.
What tools are available through the MCP interface?
The MCP server provides seventeen tools categorized into Work, Projects and documents, Files, and Accounts.
Can I use passalong with other AI clients?
Yes, it works with Codex, ChatGPT, Claude, or any client that supports MCP or HTTP.
How is work data stored?
Work is stored as plain markdown with frontmatter on the server, so nothing accumulates on your local machine.