gitagent
Your AI software factory on GitHub.
Issue in. Agents run the line. Pull request out — with you as the final station.
Add "forgot password" link to the login page #42
gitagent/fix-42→mainTrigger
Done
Triage
In progress
- 3
Coder
Pending
- 4
Reviewer
Pending
- 5
Human review
Pending
How it works
Three stations to a live line
Build the factory once. Every issue runs the same route.
- ST-01→ next
Create agents as steps
Connect the GitHub App, add Anthropic and OpenAI keys, then spin up agents from templates — triage, coder, tester, reviewer, docs — or custom prompts.
- ST-02→ next
Compose a multi-agent workflow
Chain agents in order with review gates. The workflow is what you trigger — one slug for the whole pipeline.
- ST-03
Automate from GitHub issues
Mention the workflow slug on an issue, or start a run from the dashboard. Coding agents open a pull request for your review.
Workflow builder
Compose the line once. Run it on every issue.
Chain agents on a canvas — triage, code, review — then end with human review on GitHub.
Feature delivery
workflow/fix
When an issue comment says
@gitagent-bot fix
Triage
Claude · Sonnet
Coder
Cursor · Composer 2.5
Reviewer
OpenAI · GPT-5
Workflow examples
Ready-made lines for real SDLC work
Compose agentic workflows for triage, coding, testing, review, and docs.
- LINE-A
Feature delivery
Issue-to-PR automation with coding agents, tests, and AI code review.
Triage→Coder→Tester→Reviewer→You - LINE-B
Issue triage
Autonomous triage that clarifies scope before engineering time is spent.
Triage→You - LINE-C
Bug fix pipeline
Test-first coding agents, then review loops until the PR is clean.
Tester→Coder→Reviewer→You - LINE-D
Docs automation
Keep READMEs and docs in sync with an agentic docs + review line.
Docs writer→Reviewer→You
Platform
Everything for an AI software factory
SPEC-01
Visual multi-agent workflows
Design agentic pipelines with branching, review loops, and human-in-the-loop approval steps. The workflow is what you run.
SPEC-02
Reusable AI agents as steps
Ship specialized agents with prompts, repo context, and tools — Claude, OpenAI, or Cursor (Composer and Grok), model by model.
SPEC-03
GitHub-native automation
Trigger a workflow from issues with @gitagent-bot, stream status updates, and deliver automated pull requests where your team already works.
Safe by default
Agentic AI with hard guardrails
SPEC-01
Human-in-the-loop shipping
Coding agents open PRs only. No auto-merge. No pushes to main.
SPEC-02
Isolated sandbox runs
Every agent run executes in a locked-down Docker sandbox, then tears down.
SPEC-03
Spend and runtime caps
Per-agent and per-run cost limits keep AI usage predictable.
SPEC-04
BYOK secrets security
API keys are encrypted, redacted from logs, and never exposed in the browser.
Pricing
BYOK AI agents. Free to start.
Pay Anthropic and OpenAI directly for model usage. Free only caps daily runs.
Plan
Free
Full product access with a daily run cap.
$0
- ✓Bring your own Anthropic + OpenAI keys (BYOK)
- ✓Same repositories as Paid
- ✓10 agent runs per day
- ✓Full multi-agent workflows and spend caps
Recommended
Paid
Same access as Free, without the daily run cap.
Contact us
- ✓Everything in Free
- ✓No daily run cap
- ✓Priority onboarding support
FAQ
Questions from the floor
Is gitagent only an AI coding agent?+−
No. The core product is multi-agent workflows — issue triage, coding, testing, AI code review, docs, or any custom agent roles you define and chain together.
How do I trigger a workflow?+−
Comment @gitagent-bot plus a workflow slug on a GitHub issue (for example @gitagent-bot fix), or start a run from the dashboard. Agents are the steps inside the workflow; you can also run a single agent alone.
Does it auto-merge pull requests?+−
No. Coding agents open automated pull requests only. There is no auto-merge and no push to main — you keep human-in-the-loop control.
How does BYOK pricing work?+−
Bring your own Claude, OpenAI, and Cursor API keys and pay those providers for model usage. Free includes the same repositories as Paid, with 10 runs per day. Email hello@gitagent.app for Paid (no daily run cap).
Do I need local installs for Claude, OpenAI, or Cursor?+−
No. Install the gitagent GitHub App, add your API keys, and workflow runs execute in isolated sandboxes.
Get started
Launch your AI software factory
Open the console to start, or book a demo and we will help you ship your first multi-agent GitHub workflow.