Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

Forge

Forge is an open-source, CLI-first multi-agent orchestration framework for autonomous, auditable software development. Designed for software engineers and development teams who want to incorporate AI coding agents into real codebases safely, Forge eliminates the risks of unconstrained single-model generation—such as silent bugs, architectural drift, and unverifiable claims—by organizing specialized local CLI agents into a disciplined engineering pipeline with strict separation of concerns, empirical runtime testing, adversarial diff-first review, and complete human-in-the-loop oversight.

Features

  • Multi-Role Engineering Pipeline: Coordinates specialized roles in an auditable sequence: Critic (00), Architect (01), Planner (02), Executor (03), Tester (04), Reviewer (05), and post-execution Critic (06).
  • CLI-Native Agent Invocation: Operates directly through installed developer CLI binaries (OpenCode, Google Antigravity / agy, and OpenAI Codex) as subprocesses without proprietary cloud API wrappers or token markups.
  • Empirical Black-Box Testing: Evaluates running applications through real interfaces (Headless Chromium via Playwright, REST/GraphQL APIs, CLI binaries, Library imports) under process group supervision (RuntimeSupervisor), collecting viewport screenshots, telemetry logs, and executable reproduction scripts.
  • Diff-First Adversarial Review: Change-centric quality gate where the Git diff is the primary review artifact. The Reviewer validates Executor claims against actual codebase changes and empirical test evidence before granting approval.
  • Autonomous Self-Repair Loop: In forge auto, automatically iterates between the Executor, Tester, and Reviewer. If tests fail or changes are requested, structured issue feedback is fed back into subsequent implementation attempts until approved or retry limits are reached.
  • Interactive Checkpoints & Resumption: Run interactively with step-by-step confirmation checkpoints (forge run) or resume any previous run seamlessly (--run run-XXX), automatically skipping already completed stages.
  • Project Knowledge Base (PKB): Automatically discovers and tracks persistent repository knowledge across runs in .forge/knowledge/ (architecture.yaml, features.yaml, decisions.yaml, unresolved.yaml) with role-based permissions and human locking.
  • Terminal User Interface Dashboard: Interactive terminal dashboard (forge dashboard) to monitor live runs or inspect historical runs with timeline navigation, streaming console output, stage artifacts, test evidence, and knowledge base browser.
  • Git Safety & Change Attribution: Captures unified diffs with automatic filtering of .forge/ runtime data and .env* secrets, intelligently attributing changes to avoid committing mixed-ownership files containing pre-existing user edits.
  • Process-Level Run Locking: Kernel-backed exclusive run locking (flock) prevents concurrent process conflicts on the same run directory with automatic stale-lock recovery.

Requirements

  • Python: >= 3.10
  • Git: >= 2.25 installed and available in PATH
  • Supported Agent CLIs (at least one installed and authenticated):
    • OpenCode (opencode)
    • Google Antigravity (agy or antigravity)
    • OpenAI Codex (codex)
  • Optional for Web Testing:
    • playwright (pip install playwright && playwright install chromium)

Platform Support

Forge supports Linux, WSL2, macOS, and native Windows.

Platform Status Notes
Linux Supported Primary development and reference testing platform.
WSL2 Supported Full compatibility with all pipeline stages and process supervision. Note: CLI agent tools (e.g. OpenCode) must be installed natively inside the WSL2 Linux environment; Windows host binaries exposed via WSL PATH are not supported.
macOS Supported Compatible with POSIX process isolation and standard pipeline execution.
Native Windows Supported Native Windows process trees (CREATE_NEW_PROCESS_GROUP, taskkill), file locking (msvcrt), and .cmd/.bat wrappers for agent CLIs. Windows Terminal is recommended for interactive dashboard display.

Installation

Install Forge from PyPI:

pip install forge-orchestrator

Development Installation

To contribute to Forge or install directly from the source repository:

git clone https://github.com/Mathir14/Forge.git
cd Forge
pip install -e ".[dev]"

Verify your installation:

forge --version
forge doctor

Quick Start

Get started with Forge in under 5 minutes:

# 1. Initialize Forge prompt templates and configuration in your project
forge init

# 2. Check installed CLI tools, git state, and stage configurations
forge doctor

# 3. Perform a standalone codebase audit for technical debt and vulnerabilities
forge critic "Audit authentication system and error handling"

# 4. Implement a task with step-by-step confirmation checkpoints
forge run "Add /healthz JSON endpoint with uptime and git commit SHA"

# 5. Inspect the execution artifacts and timeline in the interactive dashboard
forge dashboard

To run the pipeline completely unattended with autonomous self-repair:

forge auto "Implement rate limiting middleware using token bucket algorithm" --auto-commit

Supported Adapters

Forge connects to installed CLI tools using native adapters. View all registered adapters and their capabilities at any time using forge adapters:

Adapter Binary Default Model Prompt Transport Core Capabilities
opencode opencode Provider default stdin code_read, code_edit, shell, git, structured_output, custom_flags
antigravity (agy) agy or antigravity gemini-3.7-flash-high -p flag code_read, code_edit, shell, git, long_running, structured_output, custom_flags
codex codex gpt-5.6-terra stdin (-) code_read, code_edit, shell, git, long_running, structured_output, tool_calling, custom_flags

Documentation

For full operational details, configuration schemas, CLI command reference, testing engine guides, and troubleshooting workflows, consult the Complete User Guide.

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines on development environment setup, coding conventions, and submitting pull requests.

License

Forge is licensed under the MIT License.

Metadata

Release files for forge-orchestrator 0.1.0b9

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for forge-orchestrator 0.1.0b9
File Size Uploaded
forge_orchestrator-0.1.0b9.tar.gz 369.6 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for forge-orchestrator 0.1.0b9
File Interpreter ABI Platform
forge_orchestrator-0.1.0b9-py3-none-any.whl Python 3 none any Details

Total release size: 567.9 kB

Release files / forge_orchestrator-0.1.0b9.tar.gz

Download URL forge_orchestrator-0.1.0b9.tar.gz
Size 369.6 kB
Tags Source
SHA-256 checksum
How to use checksums
fda3a3be22827c4e19ff32b9f75ae0ef97a0f2fe2d3f0adf2c8c23c30d296ca5
BLAKE2b-256 checksum
How to use checksums
e6b1006d4c791f4534a27df4a1aabaf439fc09eebfb7c8b9aca6aa1d75c98699
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 7, 2026.

Transparency log

Release files / forge_orchestrator-0.1.0b9-py3-none-any.whl

Download URL forge_orchestrator-0.1.0b9-py3-none-any.whl
Size 198.3 kB
Tags Python 3
SHA-256 checksum
How to use checksums
d86a58efbac0f779e7285432ba36b838155284d16b26fa1039cfa29f9322e3a3
BLAKE2b-256 checksum
How to use checksums
388357e7fa6ee633d0634b6d7e66167e6a7c16f8d06b420489d11fd3da07b7cb
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 7, 2026.

Transparency log
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page