Phaedrus MCP server
Commit Nov 3, 2026Manages a curated literature-watch findings store, a research watchlist, and generates revenue-candidate drafts for integration with Lupin.
Also known as qompassai/phaedrus
- Stars
- 0
- Tools
- 7
- Downloads, 30 days
- —
- Latest version
- 1.0.0
- Last commit
- Nov 3, 2026
- License
- Apache-2.0
What is Phaedrus MCP?
Phaedrus manages a curated literature-watch findings store, a research watchlist, and generates revenue-candidate drafts for integration with Lupin. It exposes 7 tools, such as capability_context, feed_summary, finding_get.
Summary generated from the project's README
About Phaedrus MCP server
Phaedrus serves as a specialized management layer for research feeds, acting as a curated findings store and a research watchlist. It is designed for developers and researchers who need to organize literature-watch data and bridge the gap between research findings and revenue alignment. The server operates as a single Rust binary that communicates via MCP over stdio using NDJSON JSON-RPC. It functions entirely on local files within a specified data root, requiring no external database or network listener to operate.
Users interact with the server through specific tools to manage their research lifecycle. You can use finding_list to view curated findings, watchlist_list to check monitored venues and terms, and research_submit to place new candidates into a scan inbox. A unique feature is the ability to generate revenue-candidate drafts. The revenue_candidate_draft tool creates a draft payload for the Lupin server, though Phaedrus remains restricted to drafting and never writes directly to Lupin files. Setup requires configuring the command path and providing a data root via arguments or the RESEARCH_ROOT environment variable.
Written from the project's README. Every use case is checked against a quote from it.
Phaedrus MCP tools 7
Detected from the READMEcapability_contextfeed_summaryfinding_getfinding_listresearch_submitrevenue_candidate_draftwatchlist_list
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableqompassai/phaedrus is public and reachablegithub.com/qompassai/phaedrus ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-03github.com/qompassai/phaedrus ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model7 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNo tool named like delete, drop or execPass
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubPhaedrus
An MCP server for the research feed: a curated literature-watch findings store and watchlist, plus a one-way, draft-only revenue seam into lupin (the revenue-alignment server, which owns its own ledger).
Phaedrus is a single Rust binary. It speaks MCP over stdio (NDJSON JSON-RPC) to the client and works on plain files under a data root — no database, no network listener.
How it runs
- Data root (findings store, watchlist, scan inbox):
--root <path>or theRESEARCH_ROOTenvironment variable (default~/workspace/research-feed-20261019). - Lupin root (read-only — capability context and draft generation):
--lupin-root <path>orLUPIN_ROOT(default~/workspace/revenue-alignment-20261019). - Writes are deliberately narrow:
research_submitfiles a candidate into the scan inbox and never touches the curated store, andrevenue_candidate_draftonly drafts a lupinopportunity_submitpayload — this server never writes lupin's files.
Tools
| Tool | What it returns |
|---|---|
capability_context |
Lupin's capability registry (read-only) as planning vocabulary: capability ids + names, optionally keyword-filtered |
feed_summary |
Counts: findings by system/kind, systems watched, parse health |
finding_get |
One curated finding by id |
finding_list |
Curated findings, optionally filtered by system |
research_submit |
File a candidate (finding, roadmap item, or test proposal) into the scan inbox; never writes the curated store |
revenue_candidate_draft |
A draft lupin opportunity_submit payload for a revenue-shaped finding (drafts only) |
watchlist_list |
The watchlist (per-system venues and terms) |
Client config
{
"mcpServers": {
"phaedrus": {
"command": "/path/to/phaedrus",
"args": []
}
}
}
Registry
server.json in this repo is the MCP Registry record for
ai.qompass/phaedrus. The server is distributed as an MCPB bundle
attached to the GitHub releases (Linux x64).
License
Apache-2.0 — see LICENSE.
Phaedrus MCP questions
What tools does Phaedrus MCP provide?
It exposes 7 tools, including capability_context, feed_summary, finding_get, finding_list, research_submit.
Does Phaedrus require a database?
No, it works on plain files under a data root and does not use a database or network listener.
Can Phaedrus write directly to Lupin's files?
No, it only drafts a payload for Lupin; it never writes to Lupin's files.
How can I configure the data root?
You can set the data root using the --root argument or the RESEARCH_ROOT environment variable.