agent-fastpath
MCP server giving coding agents fast typed decisions (ship gates, risk checks, file triage, page checks) as accept, review or escalate.
Overview
agent-fastpath is a Jev MCP server that handles small decisions coding agents otherwise spend context on: is this CI log ready to ship, which of 40 files handle auth, did the page actually say "Order confirmed". Clear-cut cases are decided by code rules in under a millisecond; the rest go to Jev, and every result is accept, review or escalate so the agent knows when to ask a human.
Presets include ship_gate (READY_TO_SHIP, NEEDS_REVIEW, BLOCKED), risk checks for destructive commands, ambiguity detection that returns ask_user, file triage that ranks up to 500 files on the server and returns snippets (the README's benchmark shows 408 tokens read instead of 35,256), and a headless Playwright browser with SSRF protection and a gate on irreversible actions. Setup guides cover 12 MCP clients.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Is this CI or test log ready to ship? | READY_TO_SHIP · NEEDS_REVIEW · BLOCKED |
| Score | How relevant is each file to the query?(inferred) | — |
| Yes / no | Is this command risky or destructive?(inferred) | yes · no |
| Yes / no | Does the live page confirm the claim?(inferred) | yes · no |
How it works
- Agent calls a fastpath tool or preset
- Deterministic rules settle clear-cut cases
- Remaining cases go to Jev for typed, calibrated answers
- Result is mapped to accept, review or escalate
- Agent proceeds, checks, or asks the user
Strengths
- Keeps file contents out of agent context
- Rules-first keeps many decisions sub-millisecond
- Explicit uncertainty outcome tells the agent when to ask
Caveats
- Needs a TypeSafe key for the model path
- Benchmark numbers are the project's own
Getting started
Install the agent-fastpath npm package and add it as an MCP server in your client; see the setup guides on the project site.At a glance
- Type
- MCP server
- Status
- beta
- Language
- TypeScript
- Runtime
- MCP server
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Jev MCP server: a decision layer for coding agents, built on TypeSafe Jev (System One model). Ship gates, risk checks, file triage that keeps files out of context, and a safe headless browser, with calibrated confidence. For Claude Code, Codex, Cursor.
hellogumbo/awesome-jev
Frequently asked questions
What does it decide?
Ship readiness of logs, file relevance, command risk, requirement ambiguity and whether a web page confirms a claim.
How does it signal uncertainty?
Every result is accept, review or escalate.
Which clients work?
Any MCP client; guides cover Claude Code, Codex, Cursor, Gemini CLI, OpenCode, VS Code Copilot, Cline, Zed, Amp, Pi and others.
Related projects
CLI and MCP server giving coding agents Jev-powered semantic search, focused reads and list picking in about a second for about a cent.
BenchmarkbetaPythonAccessibility-first macOS computer-use MCP server and CLI, with optional Jev guards on targets and inputs before irreversible actions.
PluginbetaPythonTyped judgment layer for Claude Code, Codex and Hermes Agent: deterministic facts in code, language judgments in Jev, with receipts.
ProjectbetaPythonMCP server that takes a whole browser task (URL, goal, checks) in one call and drives Chrome server-side with a decision model.
MCP serverbetaPythonAgent skill that makes a coding agent try Jev first for bounded judgments like classify, rank or gate, and log each attempt.
Agent skillbetaPythonOrchestrator that probes which coding agents are installed and logged in, then has Jev pick the best one for a task.
Agent skillbetaPython