Tenant-aware audit log
Small API, strong tests, clear license, and materialization-ready provenance.
- Score
- 91
- Tests
- 42
- License
- MIT
Public infrastructure for reusable code
Retrieve before generate.
CodeMCP helps AI coding agents find, rank, verify, and materialize modules that already exist, then rewards the developers whose code keeps the software commons moving.
40%
minimum coder payout pool
$5
target Builder subscription
MCP
agent-native lookup and materialization
Every duplicated module is a tiny tax on software progress.
AI can generate code quickly, but every fresh copy still needs review, security checks, integration, and maintenance. CodeMCP turns prior implementations into a searchable supply of verified code shapes.
Agent lookup
Ask for a shape. Reuse the best proven module.
The local demo models the MCP request an AI coding agent should make before spending tokens on a new implementation.
Ready to query the registry.
factory-atlas materialize atlas/audit-ledger.json --capability packages/ragu-audit --out borrowed/audit
The fetched source carries `MATERIALIZED_FROM` evidence so future workers can trace owner, repo, commit, license status, and ranking.
How it works
A registry that understands shape, not just names.
Contributors seed the registry.
Developers submit open or licensable modules. CodeMCP scans the repo, fingerprints capability shapes, records provenance, and runs evidence checks.
Agents search before generating.
Codex, Claude Code, Cursor, and factory workers ask the MCP service for a needed shape and receive ranked candidates with license and safety status.
Builders materialize proven code.
The selected source slice is fetched with provenance intact, so teams know what was reused, why it ranked, and who deserves credit.
Marketplace economics
Low-margin by design. Useful enough to sustain itself.
CodeMCP is modeled as a public-service utility: tiny indexing fees, low subscriptions, transparent allocation, and a coder payout pool weighted by real reuse, quality evidence, and chart performance.
100K
$5
$500K
monthly public subscription revenue
$225K
45% monthly coder payout pool
$810K
annual top module at 30% qualified reuse
Subscriptions stay small by default. The service earns by being useful to many builders, then routes a published share to coder-artists.
Shape leaderboards
Compete on simplicity and efficacy.
Rank
Module shape
Signal
#1
Tenant-aware audit log
small API, broad tests, clean license
#2
Webhook verifier
few dependencies, fast benchmark, active maintainer
#3
RAG document chunker
high adoption, clear docs, sandbox passed
Trust layer
Discovery is not enough. Reuse needs evidence.
Provenance travels with the code.
Every materialized module carries its source repo, owner, commit, license status, and scan evidence.
Rankings are earned.
Quality scores combine shape match, tests, simplicity, safety, maintenance, and integration cost.
Public-service accounting.
Revenue allocation should be published: infrastructure, safety, payouts, salaries, reserves, and a deliberately modest surplus.
Pilot
Start with owned repos. Open the commons carefully.
The first milestone is a private atlas across real portfolios, then a small public registry with safety gates, transparent rankings, and a payout model that rewards useful code without turning the commons extractive.