Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
Open source contribution manager with PR tracking, issue discovery, and CI diagnosis
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
GITHUB_TOKENValues marked secret are keys to your account. Never share them or paste them into a chat.
You need
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Open Claude Desktop → Settings → Developer → Edit Config, and paste this into claude_desktop_config.json
{
"mcpServers": {
"oss-autopilot": {
"command": "npx",
"args": [
"-y",
"@oss-autopilot/mcp"
],
"env": {
"GITHUB_TOKEN": "your-value"
}
}
}
}Run this in your terminal
claude mcp add oss-autopilot --env GITHUB_TOKEN=your-value -- npx -y @oss-autopilot/mcpFill in the empty values before running it.
Paste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json
{
"mcpServers": {
"oss-autopilot": {
"command": "npx",
"args": [
"-y",
"@oss-autopilot/mcp"
],
"env": {
"GITHUB_TOKEN": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
Paste into .vscode/mcp.json in your project
{
"servers": {
"oss-autopilot": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@oss-autopilot/mcp"
],
"env": {
"GITHUB_TOKEN": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.
Your choice stays on this device in a cookie; no account needed.
These settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.
Runs on your computer; the tool list appears in your AI app after you install it.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 8, 2026
The registry name was verified with the GitHub account costajohnt, which owns the repo
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
This one runs on your computer; there is no internet endpoint
Runs on your computer and uses local settings instead
Runs on your computer; we can only read tool lists from internet endpoints
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
You need to fill in: GITHUB_TOKEN
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
The GitHub repo last changed –, and the latest version (5.11.0) reached the registry 6 days ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Node.js on your computer.
In the same category: Spec Driven Development, Remote Desktop Commander, DECIDE: Migration Discovery Risk Review
Servers in the same category from other makers, sorted by stars.
Reading a card:↓ Downloads per weeki★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Run after each discovery iteration to see what's still uncertain, who to ask next, and what changed.
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
MCP server for OSS Autopilot — exposes PR tracking, issue discovery, and contribution management as MCP tools for any MCP-compatible client.
| Feature | Count | Description |
|---|---|---|
| Tools | 35 | daily, status, search, features, vet, vet-list, verify-issue, track, compliance-score, repo-vet, strategy, eval, comments, post, claim, config, init, setup, check-setup, startup, dismiss, undismiss, work-claim, work-finish, work-release, work-list, move, state-show, state-sync, state-unlink, guidelines-list, guidelines-get, guidelines-store, guidelines-reset, guidelines-fetch-corpus |
| Resources | 6 | oss://status, oss://config, oss://prs, oss://prs/shelved, oss://pr/{owner}/{repo}/{number}, oss://repo/{owner}/{repo}/guidelines |
| Prompts | 4 | triage (PR prioritization), respond-to-pr (draft response), find-issues (discover issues), extract-learnings (distill per-repo guidance from past PR feedback) |
Supports stdio (default) and Streamable HTTP transports.
gh auth login)# 1. Add the server to your MCP client (see config examples below)
init tool once with your GitHub usernamedaily to check your PRs, search to find issues.
Add to your claude_desktop_config.json:
{
"mcpServers": { "oss-autopilot": { "command": "npx", "args": ["@oss-autopilot/mcp@latest"] } } }
Add to .cursor/mcp.json in your project or global config:
{
"mcpServers": { "oss-autopilot": { "command": "npx", "args": ["@oss-autopilot/mcp@latest"] } } }
codex mcp add oss -- npx @oss-autopilot/mcp@latest
Add to your Windsurf MCP config:
{
"mcpServers": { "oss-autopilot": { "command": "npx", "args": ["@oss-autopilot/mcp@latest"] } } }
Run the server in HTTP mode instead of stdio:
npx @oss-autopilot/mcp@latest --http --port 3001
The server listens at http://127.0.0.1:3001/mcp and accepts POST requests.
HTTP mode requires a bearer token on every request (stdio mode does not — it inherits the parent process identity). On first --http startup, the server generates a random 32-byte token and writes it to ~/.oss-autopilot/mcp.token with 0600 permissions. The path and "newly generated" marker are logged to stderr at startup.
Every HTTP client must send:
Authorization: Bearer <contents of ~/.oss-autopilot/mcp.token>
Requests without a valid Authorization header return 401. The Bearer scheme is case-insensitive per RFC 7235; the token value itself is compared byte-for-byte in constant time. Requests with a non-loopback Host header return 403 (DNS-rebinding defense). Requests must declare a numeric Content-Length ≤ 1 MiB — missing, non-numeric, or oversize Content-Length values return 413.
The token persists across restarts. To rotate, delete the file and restart — a new token will be generated. To relocate the token (e.g. for multi-user or CI setups), set OSS_AUTOPILOT_MCP_TOKEN_PATH to the desired absolute path before starting the server.
| Tool | Description | Read-only |
|---|---|---|
daily | Run daily PR monitoring check with prioritized summary | No |
status | Show current PR tracking status | Yes |
search | Search GitHub for contributable issues | Yes |
features | Find feature-scoped opportunities in repos with 3+ merged PRs (relationship-anchored) | Yes |
vet | Analyze an issue for contribution suitability; records the vet-time prediction and issue traits (labels, author association, competing PRs, first PR in the repo) for calibration | No |
verify-issue | Deterministically verify issue state and linked-PR claims before vetting; also returns the issue author, author association and comment count | Yes |
vet-list | Re-vet all available issues in the curated issue list | No |
track | Inspect a PR: read-only metadata lookup (nothing is tracked or persisted) | Yes |
compliance-score | Score a PR against opensource.guide best practices (#1245) | Yes |
repo-vet | Compute the repo health rubric (1–10 + verdict) for owner/repo (#1271) | Yes |
strategy | On-demand contribution strategy snapshot via computeStrategy() (#1243) | Yes |
eval | Calibration and outcome report from the outcome ledger (#1825) | Yes |
comments | Fetch and display PR comments | Yes |
post | Post a comment on an issue or PR | No |
claim | Claim an issue by posting a comment | No |
config | Get or set configuration values | No |
init | Initialize with a GitHub username | No |
setup | Configure preferences (languages, interests) | No |
check-setup | Check if setup is complete | Yes |
startup | Run startup checks (auth, state, config) | No |
dismiss | Dismiss an issue from notifications | No |
undismiss | Re-enable notifications for a dismissed issue | No |
work-claim | Claim an issue (implementation, while a fix is built in the background) or a PR (pr, while a session works it) so other sessions and machines see it; a claim another session holds needs takeover (#1856, #1852) | No |
work-finish | Mark an implementation claim finished: its branch is ready to collect for a week (session refuses a claim held elsewhere) | No |
work-release | Drop the claim on a URL, or every claim a session holds | No |
work-list | List live work claims across sessions and machines | Yes |
move | Move a PR between states (attention, waiting, shelved, auto) | No |
state-show | Show current state persistence mode (local or Gist) and sync status | Yes |
state-sync | Force push current state to the backing Gist | No |
state-unlink | Disconnect from Gist persistence and switch to local-only mode | No |
guidelines-list | List repos that have stored guidelines (always empty in local mode) | Yes |
guidelines-get | Read per-repo learning guidelines extracted from past PR feedback | Yes |
guidelines-store | Persist per-repo guidelines (8 KB cap; requires Gist mode) | No |
guidelines-reset | Tombstone the guidelines file for a repo | No |
guidelines-fetch-corpus | Fetch raw PR comment bundles for the host's extract-learnings prompt to consume | No |
| Resource URI | Description |
|---|---|
oss://status | PR tracking status (cached local state) |
oss://config | Current configuration |
oss://prs | Active open PRs from last daily digest |
oss://prs/shelved | Shelved PRs |
oss://pr/{owner}/{repo}/{number} | Detail for a specific PR |
oss://repo/{owner}/{repo}/guidelines | Per-repo learning guidelines (markdown) |
| Prompt | Args | Description |
|---|---|---|
triage | none | Fetches daily digest and builds a prioritized triage list |
respond-to-pr | prUrl | Fetches PR comments and context for drafting a response |
find-issues | maxResults? | Searches for issues ranked by viability score |
extract-learnings | repo, corpus, existingGuidelines? | Distills durable per-repo guidance from raw PR comment bundles (#867) |
With Gist persistence, every client that runs oss-autopilot (the plugin on each machine, and every MCP server process) reads and writes the same state. Since core 3.35.0, a client keeps state fields it does not recognise instead of dropping them, and warns when the state was written by a different release. Clients on older releases still drop newer fields when they save, so update every machine after a release and restart running MCP servers, which keep the version they started with.
The server can also be imported and used as a library:
import { createServer } from '@oss-autopilot/mcp';
const server = createServer(); // Connect to your own transport
See the main repository README for the full documentation, including the Claude Code plugin, CLI usage, dashboard, and contributing guide.
MIT