shipwithjev

Catalog / Agents & browsers

0152GitHub

jev-tool-runner

Jev selects the developer tools; Codex writes the code.

micic-mihajlo/jev-tool-runnerREADME ↗
# Jev native tool router

Jev selects a native Codex tool call **before the coding model is invoked**. Codex executes that call through its own tool runtime, approvals and sandbox. The result stays in Codex's conversation. The coding model handles reasoning, edits and the final answer.

The router does not open repository files, execute commands, or conduct a separate investigation. TypeSafe receives the current user goal, a small set of tool choices, and completion metadata—not source-file output.



## Run it

Requires Node.js 22+, an authenticated Codex CLI, and a TypeSafe key in a private env file. Verified with Codex CLI 0.153.4.

```sh
npm ci
npm run build

# Start a normal interactive Codex session in your project:
node /absolute/path/to/jev-tool-runner/dist/router-cli.js \
  --key-file /absolute/path/to/private.env -- \
  -C /absolute/path/to/your/project
```

Ask a task that names source files, for example:

> Read src/access.mjs and test/access.test.mjs, then explain the access check.

For an existing session, pass `resume SESSION_ID` after `--`. The wrapper supplies a local model provider for that process; it does not modify global Codex configuration or change the selected model, reasoning effort, permissions, or hook trust.

[Daily use](docs/DAILY-USE.md) · [Architecture](docs/ARCHITECTURE.md) · [Live validation](docs/evidence/native-routing.json)

## Current coverage

The Codex adapter offers bounded, numbered reads of relative source paths explicitly named in the current user message. Jev chooses among at most eight offers or hands control to the coding model. Codex executes at most four selected calls before handoff. Completed calls are excluded. Unsupported requests, missing tool support, provider failures and timeouts fall back to the ordinary coding-m

Also filed under Agents & browsers

  1. 0226

    Flight search with Browser Use★

    Breaking: Browser Use + Jev = Ultrafast ⚡ Findings flights took 7s and cost only $0.0039 🤯 > new action space every step > DOM state space > small LLM fallback to type (this video is at 1x speed btw) Built a tiny open source browser agent. try it below ↓

    @gregpr07 · Agents & browsers · ~$0.004 · ~7 s

  2. 0619

    jevmem: automatic project memory for Claude Code

    After every Claude Code message, Jev decides in ~0.3 s whether it holds a decision, rule or bug worth keeping, and jevmem saves it as one line in JEVMEM.md.

    @jetwaniavinash · Agents & browsers · ~0.3 s

  3. 0618

    Playwright + Jev: tests in plain language

    Write Playwright steps like "add the most expensive item to the cart". Jev finds the right buttons and fields, fills forms and checks the result.

    Andrey Popov · Agents & browsers

  4. 0617

    toolgate: a tool-call firewall for agents

    A Claude Code and MCP hook where Jev answers seven yes/no risk questions per tool call to allow, ask or deny. Usage and benchmark numbers are in the repo.

    Jaz · Agents & browsers