| FMRS | 64 / 100 · C | 79 / 100 · B |
|---|
| Reliability | 12 / 20 | 12 / 20 |
|---|
| Security and permissions | 9 / 20 | 18 / 20 |
|---|
| Maintenance | 14 / 20 | 16 / 20 |
|---|
| Documentation | 16 / 20 | 18 / 20 |
|---|
| Setup experience | 13 / 20 | 15 / 20 |
|---|
| Best for | - Long, multi-turn coding agent sessions
- Developers using several CLI clients at once
- Teams that need local-only operation, no telemetry, and a commercially usable license
- Teams that want to verify compression claims with a reproducible benchmark rather than vendor assertions
| - Engineering teams scaling AI coding agents while controlling spend
- Platform teams that need auditable, rollback-aware agent execution records
- Organizations using multiple agents (Claude, Codex, Gemini) under one governed flow
|
|---|
| Not for | - Anyone who wants a bare MCP server without installing client hooks or plugins (the README states adding the server alone does not enforce anything)
- Scenarios where no request content may ever be written to local disk (the proxy writes elided payloads to spill files in the OS temp directory)
- Users who need a hosted service or cloud account
- Environments that cannot run Node.js 22+
| - Developers who just want a simple code-completion helper without run governance
- Teams unwilling to accept local JSONL run records with HMAC signing
- Environments without a supported adapter (Claude/Codex/Gemini CLIs, direct provider, or verifier-only)
|
|---|
| Required permissions | - Read and write project files for smart reads, diffs and edits
- Install and run client lifecycle hooks that intercept or re-route tool calls before execution
- Write cache, knowledge graph, logs and elided-content spill files to local disk
- Optionally start a compression proxy process bound to the local loopback address
- Read native CLI usage receipts for token accounting and pricing
| - Read/write access to the configured repo root (writable scope restrictable via --allow-path/--deny-path)
- Ability to invoke the chosen agent CLI (Claude/Codex/Gemini, etc.) and its underlying model account
- Local disk writes for run records, receipts, and share outputs (share/run-receipt., etc.)
- Execution of user-configured verifier commands (e.g. npm test)
|
|---|
| Risks and side effects | - The server refuses or rewrites built-in tool calls (Read, Grep, Glob, Edit, Write, and shell cat/head/grep -r), which in enforce mode can interrupt existing workflows
- Recoverable elided content is written with 0600 permissions under token-optimizer-spill/ in the OS temp directory and is not deleted while the proxy runs; a kill signal or power loss leaves the directory behind to be cleaned up by hand
- The proxy does not inspect written content for secrets, so a secret in your conversation can reach a spill file like any other text
- The local knowledge graph and cache hold project structure, content hashes and derived findings
- Several comparisons in the README are development-stage or incomplete studies; the project itself states that superiority was not established and that aggregate intervals still span 1.0
| - Verifier commands are executed; poorly chosen verifier commands carry security risk despite preflight policy checks
- Cost and token figures may be estimates (labeled with provenance) and should not be treated as settled accounting
- Receipt integrity must be verified before use as trustworthy evidence; missing rollback evidence is flagged as EVIDENCE_BOUNDARY
- Policy defaults come from martin.config.yaml and can be overridden by CLI flags; misconfiguration can loosen safety boundaries
|
|---|
| Supported clients | Claude Code, Codex, GitHub Copilot CLI, Gemini CLI, Qwen Code, Cursor, Cline, OpenCode, Kilo, Windsurf, Roo Code, Zed, Amp, Continue, Crush, Droid | Claude Code, Codex, Gemini CLI, Cursor, VS Code, GitHub Copilot, Continue |
|---|
| Tools | 11 | 0 |
|---|