Foundry

Autonomous build loop that plans, builds, reviews, and learns — forever.

Foundry reads an IMPL_PLAN.md task list and works through it using Claude Code agents in a TUI, committing each completed task. When all tasks are done, it discovers new work and keeps going.

Demos

• Building a Second Brain with the Loop — Foundry autonomously works through an implementation plan, building a second-brain app from a task list while the TUI streams each agent's output in real time.
• Enhancing the Second Brain with the Loop — A follow-up run where foundry picks up where it left off, discovering new work and iterating on the second-brain app with patterns learned from the first pass.

Task Flow

Load patterns from ~/.foundry/patterns/
  │
PLANNER (+ matched patterns) → .buildloop/current-plan.md
  │
BUILDER → implement plan
  │
REVIEWER (+ patterns + runtime checks) → .buildloop/review-report.md
  │  └─ if FAIL: FIXER → re-review (up to 2 passes)
  │
PATTERN EXTRACTOR → merge into ~/.foundry/patterns/
  │
GIT COMMIT → feat(task_id) or WIP(task_id)

How It Works

Foundry is a harness for Claude Code. Each agent (planner, builder, reviewer, fixer, discoverer) is a Claude Code CLI invocation with a role-specific prompt and scoped tool access. The Rust binary handles orchestration, streaming, and state — Claude does all the reasoning and file editing.

The loop

Without guardrails, an autonomous build loop degrades fast. Task 3 builds on task 2's mistakes, which built on task 1's mistakes. Errors compound and the codebase drifts from the intended architecture.

Foundry's loop is designed around two forms of backpressure:

Short-term: the review gate. After the builder finishes, a reviewer agent audits the changes — running build checks, linting, tests, and a structured code audit. If the reviewer finds HIGH or MEDIUM severity issues, a fixer agent addresses them, and the reviewer runs again. This happens up to 2 passes. If the work passes, it gets a feat(task-id) commit. If it still fails after 2 passes, it gets a WIP(task-id) commit — a signal that the work isn't validated and shouldn't be built on confidently. The review gate prevents bad code from silently flowing forward.

Long-term: pattern learning. After each validated task, a pattern extractor agent scans the build artifacts, review findings, and plan to extract reusable lessons (e.g., "CFrame not Position for moving Roblox parts" or "always validate UTF-8 boundaries before string slicing"). These get saved as structured JSON to ~/.foundry/patterns/. On the next task — in any project — matched patterns are injected into the planner and reviewer prompts as reference data. Patterns that recur 3+ times get auto-promoted, meaning they're always included. This is how the system gets better over time: a mistake made once becomes a check applied everywhere.

Pattern scope

Patterns are global by default. They live in ~/.foundry/patterns/ and are loaded for every project on your machine. A lesson learned building project A is available when building project B.

If you want per-project isolation, set patterns_dir in .foundry.json to a project-local path.

Discovery

When all tasks in IMPL_PLAN.md are complete, foundry doesn't stop. A discovery agent scans the codebase — reading architecture docs, looking for TODOs/FIXMEs, checking for failed tests, spotting inconsistencies — and appends new tasks to IMPL_PLAN.md. The loop then works through those. If discovery finds nothing, it sleeps and tries again later.

Install

# From source (requires Rust toolchain + Claude Code CLI)
cargo install --git https://github.com/context-foundry/context-foundry foundry

Or via Homebrew (once a release is published):

brew tap context-foundry/tap
brew install foundry

Prebuilt GitHub Release binaries are published for macOS, Linux, and Windows x86_64. foundry update uses those release assets directly.

Or build locally:

cargo build --release
# Binary at ./target/release/foundry

Usage

Point foundry at any project directory that has an IMPL_PLAN.md:

# TUI mode (default)
foundry --dir /path/to/project

# Interactive prompt-driven studio for Claude, Codex, or both
foundry --dir /path/to/project studio

# Headless mode (CI/logs)
foundry --dir /path/to/project run --no-tui

# Check progress
foundry --dir /path/to/project status

# List all tasks
foundry --dir /path/to/project tasks

# Self-update to latest release
foundry update

Studio is documented separately in docs/foundry-studio-readme.md because it is a different workflow from the autonomous run loop.

Project Setup

A project needs two files to get started:

1. IMPL_PLAN.md — Task checklist (foundry reads and marks tasks done):

   ## Phase 1
   - [ ] 1.1: Set up project scaffolding
   - [ ] 1.2: Implement authentication
   

2. ARCHITECTURE.md — What you're building (agents read this for context)

Optional:

• .foundry.json — Override defaults:

  {
    "planner_model": "opus",
    "builder_model": "sonnet",
    "reviewer_model": "opus",
    "fixer_model": "opus",
    "patterns_dir": "~/.foundry/patterns"
  }
  

• CLAUDE.md — Project conventions (agents read this too)

Agent Prompts

All agent prompts are defined in src/prompts.rs. Each agent has a dedicated prompt function:

Agent	Function	Purpose
Planner	planner_prompt()	Creates implementation plans from task descriptions
Builder	builder_prompt()	Implements the plan, runs stack-appropriate build checks
Reviewer	reviewer_prompt()	Combined validation + audit with structured findings
Fixer	fixer_prompt()	Fixes HIGH/MEDIUM issues from the review report
Discovery	discovery_prompt()	Scans the codebase for new tasks
Pattern Extractor	pattern_extraction_prompt()	Extracts reusable patterns from completed work

Prompts are compiled into the binary. To customize them, edit src/prompts.rs and rebuild.

Key design decisions in the prompt system:

• Stack-aware: agents detect the tech stack from repo files (Cargo.toml, package.json, pyproject.toml) rather than assuming a specific language
• Safe by default: the reviewer only runs read-only checks (no docker compose up, no service mutations)
• Pattern isolation: learned patterns are injected as clearly delimited reference data, not as authoritative instructions
• Evidence-based review: every finding must cite file, line number, and concrete evidence

Extensions

Foundry does not ship with any extensions. Extensions are something you create yourself — domain-specific knowledge packages that teach foundry's agents how to work with your particular technologies, APIs, or workflows.

An extension is just a folder containing guides, patterns, templates, and examples (e.g., a CLAUDE.md with rules, a patterns JSON with learned issues). Foundry provides the framework: agents already know how to read CLAUDE.md files for project conventions, and the pattern system already knows how to load and match JSON pattern files. You supply the domain knowledge.

To use an extension, copy its CLAUDE.md and relevant docs into your project, or reference them from your project's CLAUDE.md.

Architecture

• config.rs — Settings with serde defaults (backward-compatible JSON)
• agent.rs — Spawns Claude CLI in a PTY for real-time streaming
• patterns.rs — Load, match, format, merge, and extract learned patterns
• prompts.rs — Agent prompts (planner, builder, reviewer, fixer, discovery, pattern extractor)
• studio/ — Prompt-driven multi-model TUI with workspace isolation, artifact capture, and modular Studio app/state/UI code
• update.rs — Self-update from GitHub Releases with checksum verification
• app.rs — Build loop orchestration, review loop, pattern extraction
• tui.rs — Ratatui terminal UI with live agent output
• task.rs — Parse IMPL_PLAN.md task lists
• git.rs — Commit and push helpers

Previous Version

The Python MCP server + daemon that preceded this Rust rewrite is archived at:

• Tag: v1.0-python
• Branch: archive/python-mcp