Agentfork

Agentfork is an open source developer tool that forks LLM agents at any step and sweeps variants to test fixes.

Agentfork is an open source tool designed for developers who build and test LLM agents. It records every LLM and tool call locally, storing traces in a SQLite database without sending data to external servers. When an agent fails at a specific step, developers can fork the execution from that exact point, replay earlier steps from cache, and test changes to prompts, models, or tool results. The tool also supports parallel sweeps to test multiple variants repeatedly and generate pass rates, token usage, time, and cost metrics. Additionally, developers can save runs as pytest tests for CI pipelines or use the local web UI to browse, fork, sweep, and compare runs visually.

Agentfork at a glance

Best for: Software engineer
Runs on: self-hosted, open source, CLI, Web
Pricing: Free
Released: October 2026
Category: AI coding tool
Website: github.com/bimal1023/agentfork

Key features

Record LLM and tool calls locally into traces database
Fork agent execution from any step
Sweep variants with repeats and parallel execution
Local web UI to browse and compare runs
Save runs as pytest tests for CI
Compare two runs step by step
Estimate costs per variant based on model list prices

Use it for

Test prompt changes without re-running earlier steps
Compare pass rates across different models
Save failing agent runs as automated pytest tests
Analyze token usage, duration, and cost per variant

Companions

Tools that articles about Agentfork recommend using with it, each with its job.

Context7 — Provides up-to-date documentation for AI agents and code editors.
Ollama — Run local models at zero marginal cost, with code kept on your machine.
OpenRouter — Access over 500 models and 80 providers through one unified API.
Daytona — Runs sessions in isolated sandboxes synced to git branches.
Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
Oh My OpenCode Slim — Lightweight multi-agent orchestration and background tasks.

Alternatives to Agentfork

AgentForge — It provides a flexible architecture for building specialized AI agents.
OpenCode — Works with 75+ model providers, local models, and existing subscriptions
Agent Trials — Test if an agent fixes your bugs in your stack at a sane price before deployment.
Stagewise — Maintain full sovereignty over AI models while running agents in parallel.
fast-agent — Improves agent accuracy and cost efficiency via MCP integration.
AgentsView — It keeps your AI coding session archive local by default while tracking costs and activity.
pstack — Improves AI coding agent outcomes with rigorous workflows.
AGENTS.md — It separates agent-specific instructions from human-focused README files.
Open Agents Template — Spawn autonomous coding agents with full sandboxes and durable workflows.
AgentOps — Track token usage, costs, and traces across 400 plus LLMs and agent frameworks.
OpenMonoAgent — Runs locally on your own hardware with no recurring subscription fees.
Stoneforge — Runs isolated git worktrees and automated merges to prevent conflicts.

All Agentfork alternatives