replen — Claude Code MCP server: what it does, how to install
replen connects an AI coding agent to open-source projects that fit the repository in view. It helps review candidates, keep decisions, and create handoff PRs.
claude mcp add --env DIGEST_BASE_URL=<DIGEST_BASE_URL> --env DIGEST_TOKEN=<DIGEST_TOKEN> --transport stdio mcp -- npx -y @replen/mcpWhat replen does
replen exposes 15 MCP tools for finding and tracking project candidates. The main tool, replen_match, reads the current repository context and returns matching projects with metadata, similarity scores, and selection notes. The agent can then record a verdict, change a candidate’s state, or request a handoff PR.
Its parts cover matching, portfolio memory, and project onboarding. Atlas keeps past verdicts, reports, notes, cross-project links, queues, and saved carts. Onboarding stores grounded capabilities, domain tags, pinned runtime and dependency versions, plus product grouping. A DIGEST_TOKEN is required; without it, the MCP server cannot connect.
The setup command registers replen through npx, backs up the existing Claude Code configuration, and leaves other MCP servers intact. Candidate triage is handled by the host agent through its own subscription, while Replen uses embeddings and a catalogue classifier on its side.
What it helps with
- find open-source projects that fit the current repository
- record a verdict for each reviewed candidate
- recall decisions and notes across a project portfolio
- open a handoff PR for a selected match
Example prompts for the agent
- Find new tools that fit the repository I have open and explain why they were selected.
- Use
replen_match, review the candidates against this codebase, then save each verdict withreplen_record_triage. - Create handoff PRs for every project I have starred.
How to install replen
Platform CLI
claude mcp add --env DIGEST_BASE_URL=<DIGEST_BASE_URL> --env DIGEST_TOKEN=<DIGEST_TOKEN> --transport stdio mcp -- npx -y @replen/mcpReplace <…> with your own values
Platform CLI
codex mcp add mcp --env DIGEST_BASE_URL=<DIGEST_BASE_URL> --env DIGEST_TOKEN=<DIGEST_TOKEN> -- npx -y @replen/mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"mcp": {
"command": "npx",
"args": [
"-y",
"@replen/mcp"
],
"env": {
"DIGEST_BASE_URL": "<DIGEST_BASE_URL>",
"DIGEST_TOKEN": "<DIGEST_TOKEN>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@replen/mcp"
],
"env": {
"DIGEST_BASE_URL": "<DIGEST_BASE_URL>",
"DIGEST_TOKEN": "<DIGEST_TOKEN>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e DIGEST_BASE_URL=<DIGEST_BASE_URL> -e DIGEST_TOKEN=<DIGEST_TOKEN> mcp npx -y @replen/mcpReplace <…> with your own values
Step 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 replen
How do I install replen in Claude Code?
Get a token from the replen settings page, then run claude mcp add replen -e DIGEST_TOKEN=<DIGEST_TOKEN> -- npx -y @replen/mcp. Restart Claude Code afterward.
What does replen do?
It matches the repository in view with open-source projects, stores the agent’s decisions, and can open handoff PRs. It also keeps repository capabilities, tags, versions, and portfolio memory.
What does replen need to run?
It needs DIGEST_TOKEN supplied as an environment variable. The host agent handles candidate reasoning in its own session, while replen provides matching and record-keeping tools.
