Skip to main content
AgentForge Logo

AgentForge Toolkit (af-toolkit)

Fast AF developer tooling for AI agents.

Build, test, slice, audit, evaluate, and sandbox Model Context Protocol (MCP) servers and agent trajectories.

Python 3.10+ License: MIT Code style: ruff


Key Modules & Capabilities

+-------------------------------------------------------------------------+
|                           AgentForge Toolkit                            |
+-------------------+-------------------+---------------------------------+
| Forge             | Slice             | Audit & Observability           |
| - OpenAPI to MCP  | - AST Skeletons   | - Trajectory Diagrams & Diffs   |
| - Mock Servers    | - Repo Slicer     | - Safety & Policy Scanner (A-F) |
| - Python to MCP   | - Git Diff Slicer | - Benchmark Comparator          |
| - Docker Exporter | - Multi-Model Tok | - Cost Estimator ($)            |
| - Client Exporter | - Prompt Wrappers | - PII Redactor & Failure Diag   |
| - Compliance Test | - JSON & MD Slicer| - Commit Message Generator      |
+-------------------+-------------------+---------------------------------+
| Gateway & REPL    | Eval & Sandbox    | Studio & DX                     |
| - Multi-Server GW | - SWE-bench Eval  | - Interactive Web Workbench     |
| - Stdio Debugger  | - Git Snapshot    | - System Doctor & Initializer   |
| - Polyglot Bridge | - Auto-Rollback   | - Antigravity Skill Exporter    |
| - Interactive REPL| - Resource Bounds | - IDE Agent Rules Exporter      |
+-------------------+-------------------+---------------------------------+

Quickstart

1. Installation

pip install af-toolkit

2. Generate an MCP Server from OpenAPI

af forge generate api_spec.json -o server.py --service github

3. Generate Client Configuration

af forge config server.py --client all

4. Slice Code & Repositories for LLM Prompts

af slice-repo ./my_project --budget 6000 -o repomap.md
af tokens ./my_project/main.py

5. Audit Trajectory Logs

af audit report trajectory.jsonl -o audit_report.md
af audit safety trajectory.jsonl
af audit cost trajectory.jsonl --model claude-3-5-sonnet

6. Interactive MCP Debugger & REPL

af mcp-repl "python server.py"

7. Run Agent Benchmark Evaluations

af eval run tasks.jsonl --agent-cmd "agy" -o eval_report.md

8. Ephemeral Sandbox with Auto-Rollback

af sandbox run "pytest tests/" --rollback

9. Launch AgentForge Studio

af studio --port 8000

CLI Command Reference

Command Description
af forge generate <spec> Generate FastMCP server from OpenAPI 3.0/3.1 or Swagger 2.0.
af forge mock <spec> Generate offline mock MCP server with synthetic responses.
af forge python <source.py> Convert Python functions directly into MCP tools.
af forge doc <server.py> Generate Markdown documentation tool catalog.
af forge docker <server.py> Generate Dockerfile and docker-compose.yml.
af forge config <server.py> Export ready-to-use JSON config for Claude Desktop / Cursor / Antigravity.
af test-mcp <server_or_spec> Validate MCP schemas, signatures, and docstrings.
af slice <file.py> AST code skeletonization under token budgets.
af slice-repo <dir> Repository architecture map for system prompts.
af tokens <target> Multi-model token calculator across GPT-4o, Claude, Gemini, DeepSeek.
af slice-json <file> Prune large JSON/YAML data files.
af slice-md <file> Extract Markdown headings and lead sections.
af audit report <log> Trajectory metrics, Mermaid sequence diagrams, and git diff snippets.
af audit safety <log> Security scan for leaked keys, destructive commands, loop storms.
af audit compare <l1> <l2> Side-by-side benchmark comparison between agent runs.
af audit cost <log> Calculate dollar cost ($) across LLM models.
af audit redact <log> Sanitize PII, credentials, and IPs in trajectory logs.
af audit diagnose <log> Cluster tool errors and generate parameter suggestions.
af audit commit-msg <log> Generate conventional commit message from touched files.
af mcp-repl <server_cmd> Interactive stdio terminal REPL for debugging MCP servers.
af gateway --servers <cmds> Multi-server MCP multiplexer gateway.
af eval run <tasks.jsonl> Run agent evaluation harness and generate scorecards.
af sandbox run <command> Execute command inside snapshot-guarded sandbox.
af sandbox snapshot/rollback Manage git working directory snapshots.
af export-skill <server.py> Export Antigravity skill package (SKILL.md, scripts/, references/).
af export-rules Export .cursorrules, .windsurfrules, or AGENTS.md.
af config set/list/path Manage credentials in ~/.af/auth.toml.
af config export-env Export credentials to .env.
af doctor System & environment diagnostics.
af init [--ci] Scaffold project and CI/CD workflows.
af studio Launch FastAPI web workbench.

License

MIT License. Created for AI agent developers.

Metadata

Release files for af-toolkit 0.1.0

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

Source distribution (sdist)

Source distribution for af-toolkit 0.1.0
File Size Uploaded Provenance
af_toolkit-0.1.0.tar.gz 42.6 kB With provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for af-toolkit 0.1.0
File Interpreter ABI Platform
af_toolkit-0.1.0-py3-none-any.whl Python 3 none any Details

Total release size: 101.3 kB

Release files / af_toolkit-0.1.0.tar.gz

Download URL af_toolkit-0.1.0.tar.gz
Size 42.6 kB
Tags Source
SHA-256 checksum
How to use checksums
07b604d4292a175abed86fb7be62cd1cb2a80e0f6366b0c4ae1b4bddfb0cc5ee
BLAKE2b-256 checksum
How to use checksums
acc7a56647bc591fa5f2d0ec4b670eb1c54efc5e308c60a5d7e69a755cc33f14
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 Aug 16, 2026.

Transparency log

Release files / af_toolkit-0.1.0-py3-none-any.whl

Download URL af_toolkit-0.1.0-py3-none-any.whl
Size 58.7 kB
Tags Python 3
SHA-256 checksum
How to use checksums
8a0c2c7470fd3cba85e2043a1fa3526577688c04396682fa4422717d15eb048d
BLAKE2b-256 checksum
How to use checksums
9629f00d1ec3a4bd6e46dd4735c888f0cb2398fdc889df93b1f626f4a2e6a527
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 Aug 16, 2026.

Transparency log

Security for af-toolkit 0.1.0

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.

Full provenance

All 2 files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for af_toolkit-0.1.0.tar.gz

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 Aug 16, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files Full provenance

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