Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/zoeynine/local-codex-bridge/agents-mdgit clone --depth 1 https://github.com/zoeynine/Local-Codex-BridgeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/zoeynine/local-codex-bridge/agents-md)<a href="https://agentmods.dev/instructions/zoeynine/local-codex-bridge/agents-md"><img src="https://agentmods.dev/badge/instructions/zoeynine/local-codex-bridge/agents-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01130 | $0.01130 |
| Opus 5 | $0.00565 | $0.00565 |
| Sonnet 5 | $0.00226 | $0.00226 |
| Haiku 4.5 | $0.00113 | $0.00113 |
Grade A, and why
Local-Codex-Bridge AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Agent Guidance
Project intent
Local Codex Bridge is a thin MCP stdio control surface for native Codex sessions, with a shared Windows/macOS core and optional platform-specific surfaces. Its purpose is to let ChatGPT or another MCP client supervise official Codex threads without creating a second task system, transcript store, queue, retry loop, or authority layer.
Keep the bridge thin. Native Codex owns persistent threads, turns, history, final messages, and execution capabilities. Bridge-owned state is limited to bounded live supervision data, pending requests, terminal snapshots, optional bounded checkpoints, and the optional UX projection.
Architecture
The runtime chain is:
MCP client -> Local Codex Bridge (JSON-RPC stdio)
-> official `codex app-server --listen stdio://` (JSONL stdio)
-> native Codex sessions
src/mcp.ts owns the MCP boundary, src/app-server.ts owns the official child-process protocol, src/tools.ts owns the public tool contract, and src/runtime.ts owns ephemeral live state. src/checkpoint.ts provides the separate optional local checkpoint store. The Windows Tray and Secure MCP Tunnel integration are optional layers; the Tunnel itself is external to this repository.
The eight public tools have distinct semantics:
codex_threads: list/search/read persistent native threads; filters are not access control.codex_models: read one bounded currentmodel/listpage on demand; it creates no catalog cache or current-model registry.codex_turn: create or resume a native thread and start a turn; acceptance is not completion.codex_observe: read bounded live state or explicitly degraded persisted history after Bridge state loss.codex_steer: append a semantic correction to the exact active turn; do not use it as a timer or retry.codex_respond: answer one real pending app-server request using its raw ID and exact scope.codex_interrupt: interrupt one exact native turn; it is not process control.codex_checkpoint: maintain optional bounded supervisor cognition metadata; it is not a transcript or lifecycle database.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 67 lines · 1,130 tokens per session scan A 234aa78fd8ea
Local-Codex-Bridge AGENTS.md is an instructions file published in the GitHub repository zoeynine/Local-Codex-Bridge (23 stars, last pushed 10d ago), licensed MIT. It adds 1,130 tokens to every session, about $0.0056 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
douyin-transcript-mcp AGENTS.md
AGENTS.md instructions for mei-shui-xing/douyin-transcript-mcp, covering instructions for ai coding agents, required workflow and non-negotiable boundaries.
enquire-mcp AGENTS.md
AGENTS.md instructions for oomkapwn/enquire-mcp, covering agents.md, tl;dr, architecture (5-minute orientation), conventions and typescript.
mcp-app-typescript-template AGENTS.md
AGENTS.md instructions for pomerium/mcp-app-typescript-template, covering repository guidelines, project overview, development commands, quick start and all available commands.
ThreeBrowserStudio AGENTS.md
AGENTS.md instructions for SamG-Coder/ThreeBrowserStudio: ThreeBrowser Studio is an LLM-first authoring runtime. MCP is the editor and the native WebGPU window is the shared live viewport. Preserve that product shape: do not grow a conventional inspector-heavy browser UI.
PatchWarden AGENTS.md
Instructions for jiezeng2004-design/PatchWarden, covering patchwarden agent rules, commands, safety contracts, changes and release and codex memory.
sdk CLAUDE.md
Claude Code instructions for WaniWani-AI/sdk, covering waniwani sdk, public api tiers, oss (no api key required), free tier (one env var: waniwaniapikey) and internal (not part of the public api).