PAPI — Claude Code MCP server: plan → build → review cycles
PAPI: Persistent Adaptive Project Intelligence is an MCP server that guides projects through plan → build → review → release cycles. The agent knows the task stage and next step.
MIT ★ 12 updated 2026-10-15
claude mcp add --transport http papi https://mcp.getpapi.ai/mcpWhat PAPI: Persistent Adaptive Project Intelligence does
PAPI turns chat chaos into cycle structure. The agent plans right-sized tasks, records what was built and what changed along the way, and closes the loop with review. Every new session starts where the previous one left off: the agent sees the current cycle, open tasks, and the next step.
Connect via the remote endpoint https://mcp.getpapi.ai/mcp or the local @papi-ai/server package. You need a getpapi.ai account and a connection token, or browser sign-in. The web dashboard shows cycles, board, and decision log: you see project state without asking the agent.
Free tier gives the full plan → build → review → release loop on three projects. No card, not a trial.
What it helps with
- hold project direction across agent sessions
- persist chat decisions into a structured log
- see current cycle and task board without querying the agent
- get plans made of tasks the agent executes directly
- close review and release so the next cycle starts with context
Example prompts for the agent
- Read
https://getpapi.ai/llms.txtand set up PAPI - Run
orientand show where the current cycle stands - Plan the next cycle and break tasks into build steps
How to install PAPI: Persistent Adaptive Project Intelligence
1Platform CLI: remote server
claude mcp add --transport http papi https://mcp.getpapi.ai/mcp2Platform CLI
claude mcp add papi-persistent-adaptive-project-intelligence -- npx -y @papi-ai/server1Platform CLI: remote server
codex mcp add papi-persistent-adaptive-project-intelligence --url https://mcp.getpapi.ai/mcp2Platform CLI
codex mcp add papi-persistent-adaptive-project-intelligence -- npx -y @papi-ai/serverConfig file
{
"mcpServers": {
"papi-persistent-adaptive-project-intelligence": {
"url": "https://mcp.getpapi.ai/mcp"
}
}
}{
"mcpServers": {
"papi-persistent-adaptive-project-intelligence": {
"command": "npx",
"args": [
"-y",
"@papi-ai/server"
]
}
}
}Config file
{
"servers": {
"papi-persistent-adaptive-project-intelligence": {
"type": "http",
"url": "https://mcp.getpapi.ai/mcp"
}
}
}{
"servers": {
"papi-persistent-adaptive-project-intelligence": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@papi-ai/server"
]
}
}
}1Platform CLI: remote server
gemini mcp add -t http papi-persistent-adaptive-project-intelligence https://mcp.getpapi.ai/mcp2Platform CLI
gemini mcp add papi-persistent-adaptive-project-intelligence npx -y @papi-ai/serverStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about PAPI: Persistent Adaptive Project Intelligence
How to install PAPI in Claude Code?
Run claude mcp add --transport http papi https://mcp.getpapi.ai/mcp, then open /mcp → **papi** → **Authenticate** and sign in via browser, or add the token in headers when connecting.
What does PAPI do?
PAPI gives the agent a planning, build, review, and release cycle. The agent writes to the task board and decision log while working, so every new session starts with the context left by the previous one.
How is PAPI different from Linear or Jira?
Linear and Jira were built for humans who read and write tickets in a separate tab. PAPI records state as a side effect of the agent working: starting a build opens the task, finishing writes the report, release closes the cycle.