Forge is an agentic platform where AI agents plan, code, test, and open real GitHub pull requests from your tickets — with per-step cost estimates before anything runs. You approve the plan, review the PR, and merge. Alternative to Cursor, Copilot, and Devin.
Every mission flows through a configurable pipeline of specialized agents. Enable or disable any step — skip docs, skip testing, or run the full suite.
Click any step to enable / disable it
Breaks the task into sub-steps, identifies files, and builds the execution plan.
Implements the plan — writes code, creates files, and iterates until the build passes.
Writes and runs tests. Fixes failures automatically before handing off.
Code review, linting, CI checks. Pushes fixes and opens the PR when ready.
Generates or updates documentation, changelogs, and inline comments.
The New Mission modal lets you set type, priority, labels, repo, budget cap — and toggle each pipeline step on or off before launch.
Forge estimates costs for each agent role upfront — broken down by pipeline step. Track spend in real time and set budget caps to stay in control.
See cost per agent role before approving. Know exactly where your budget goes — planning, coding, testing, review, or docs.
Watch costs accumulate token by token. Every API call is logged with its exact price per agent.
Set monthly budgets and per-mission limits. Agents pause automatically when thresholds are hit.
Estimates improve over time. Forge uses your historical data to predict costs by complexity tier.
Bring your own API keys. Forge adds zero markup on model costs.
Web dashboard, CLI, and mobile app — all connected to the same agent pipeline.
Full-featured GUI with real-time agent traces, kanban boards, CI integration, and the Intelligence Hub for model management and cost tracking.
Everything the GUI does, from your terminal. Create missions, approve plans, watch agent traces, and manage budgets — without leaving your editor.
Review and approve from anywhere. Resume tasks, check costs, and manage your agents on the go — all from your phone.
The Forge CLI gives you full parity with the web dashboard. Create missions, stream agent output, approve plans, and track spend — all from the command line.
The Forge mobile app lets you stay connected to your agents wherever you are. Get push notifications when a plan needs approval, review diffs, check costs, and resume paused tasks — all from your phone.
From ticket to merged PR in under 5 minutes — fully autonomous.
Forge is a complete AI engineering platform — not just a code autocomplete. Every feature is designed for real production workflows.
Plan, code, test, fix — fully automated. Self-healing agents iterate until CI passes.
Branches pushed, PRs opened with full descriptions, merge gates enforced — zero manual steps.
Real-time GitHub Actions monitoring. Agents wait for CI, fix failures, and re-push automatically.
Get notified when agents need input or finish work — via Slack, email, or webhooks.
Set OKRs, auto-generate missions from goals, and track progress across repositories.
Full cost analytics, model comparison, usage audit logs, and budget forecasting.
Use Claude, Gemini, or local models via Ollama. Assign different models to different agent roles.
GitHub OAuth, API tokens, role-based access. Every action scoped to your organization.
Watch agents think and work in real time. Live diffs, tool calls, and reasoning — via WebSocket.
Forge is an agentic AI platform where AI agents autonomously turn your tickets into merged pull requests. You describe work in plain language, and Forge's agent pipeline reads your codebase, proposes a step-by-step plan with per-agent cost estimates, and — after your approval — implements the code in an isolated git worktree, runs tests, opens a GitHub pull request, and waits for CI to pass. Build and run with AI agents — you review and merge.
Cursor and GitHub Copilot are code assistants — they help you write code in your editor, but you still handle planning, testing, CI, and pull request creation. Devin is a fully autonomous agent but offers limited cost transparency. Forge is fully autonomous (ticket to merged PR) with mandatory human approval gates, per-step cost breakdowns before execution, a configurable 5-agent pipeline, and budget caps. You bring your own API keys with zero markup on model costs.
Most tickets cost $0.08–$1.20 in LLM API costs. Complex features with 20+ steps typically cost $2–5. Forge shows you the per-agent cost breakdown (Planning, Coding, Testing, Review, Docs) before anything runs, so there are never surprises. You bring your own API keys and Forge adds zero markup.
Forge supports Claude (Sonnet 4.6, Opus 4.6), Google Gemini (3 Flash, 3.1 Pro, 2.5 Flash), and local models via Ollama. You can assign different models to different pipeline roles — for example, Gemini Flash for planning (cheaper) and Claude Sonnet for coding (better quality). Models are swappable per mission or per agent role.
Yes. Every mission runs in an isolated git worktree — the agent never touches your main branch directly. Nothing is pushed without human approval. The output is a standard GitHub PR that goes through your normal code review and CI pipeline. Role-based access control (Viewer, Engineer, Admin) ensures only authorized users can approve and merge.
Forge uses a configurable 5-agent pipeline: Planning (Atlas-7) reads your codebase and creates an execution plan, Coding (Helix-1) implements the plan, Testing (Nova-3) writes and runs tests, Review (Echo-5) performs code review and opens the PR, and Documentation (Prism-2) generates docs. Each step can be toggled on or off per mission, and the cost estimate updates in real-time as you configure it.
Yes. The Forge CLI has full feature parity with the web dashboard. You can create missions, stream real-time agent traces, approve plans, check budget usage, and manage your pipeline — all without leaving your terminal. A mobile app is also coming soon for approvals on the go.
Join engineering teams already using Forge to clear their backlogs and ship production-quality code autonomously.