pi-firecode

A modular Pi extension for terminal UI, session workflows, adversarial review, and delegated agents

Packages

Package details

extension

Install pi-firecode from npm and Pi will load the resources declared by the package manifest.

npm repo home report

$ pi install npm:pi-firecode

Package: pi-firecode
Version: 1.6.0
Published: Oct 16, 2026
Downloads: 1,647/mo · 1,525/wk
Author: surgee
License: MIT
Types: extension
Size: 448 KB
Dependencies: 0 dependencies · 4 peers

Pi manifest JSON

{
  "image": "https://raw.githubusercontent.com/Suge8/firecode/main/design/promo/og-1280x640.png",
  "video": "https://raw.githubusercontent.com/Suge8/firecode/main/design/promo/hero.mp4",
  "extensions": [
    "./dist/index.js"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

Why FireCode

Works out of the box. The recommended config is written on first launch; swap in the models you are logged in to. The UI and the prompts follow your system language.
Sub-agents under control. Run them in parallel, pick a model per role, switch automatically on failure, step in any time.
Hands-off review. Complex work is reviewed by several models and sent back to be fixed until it passes. You see how each review ended.
An ultra-light core. Built the Pi way: minimal, so your context stays lean.
Low visual load. Focus on the results that matter. Everything in between folds away until you open it.

Fully automatic review and delivery

Just tell the commander what you want. On complex tasks it starts an adversarial review on its own: any number of models you choose look for problems, every problem goes back to be fixed, and the fix is reviewed again. It stops when the review passes, an advisor model calls it after repeated failures, or the rounds run out. You can also run /fire-review yourself at any time.

Other reviewers mostly hand you a list of problems. FireCode keeps fixing and re-reviewing:

	Reviewers	What you get
FireCode	Any number of models, reviewing adversarially	Fixed code, re-reviewed
Claude Code `/code-review`	One sub-agent of your session	A findings list; `--fix` applies it
Codex `/review`	One reviewer on your session's model	A summary of the issues
Open Code Review (Alibaba)	One model, a sub-agent per file group	Line comments to fix yourself
Qodo	Specialist agents plus a judge	Ranked findings; can apply a fix
PR-Agent (open source)	One model call per tool	Comments and code suggestions
multi-model-review (Pi)	Several model families	One combined verdict

Checked against each product's docs on 2026-10-12.

Clear sub-agent status

Every sub-agent sits above the input box. Click one to read its whole run and talk to it directly.

An ultra-light core

Built the Pi way: sub-agents add just two small tools, and review and the watcher add none. Here is what sub-agent tools cost on every request:

	Sub-agent tools	Characters
FireCode	2	1,713
pi-subagents (0.77.0)	2–3	1,561–4,918
omp (18.10.1)	2	3,369
Codex CLI (0.162.1)	5–6	5,045–9,641

Tool definitions sent to the model (name, description and parameters), default settings, measured 2026-10-12.

Also in the box

/fire-watch a cheap model that speaks up only when work drifts · /preset switch model, thinking and tools in one go · /quota and /tokens usage at a glance · Ctrl+R rename the session · Ctrl+Shift+S OpenAI fast mode · OpenAI native compaction · Works with Claude subscription login · Claude screenshots upload once instead of every turn

Install

pi install npm:pi-firecode

Restart Pi (1.3.0+) and it's ready. The config is created on first launch at ~/.pi/agent/extensions/firecode/config.jsonc: put in the models you use, and turn anything off with "features": { "<name>": false }.

Using herdr? Run herdr integration uninstall pi first; FireCode reports the pane state itself.
Sub-agents and reviewers can run bash without a sandbox. Run Pi in a container if you need isolation.