LegionCode
The open-source AI coding agent.
Done — I fixed the dashboard hero layout and updated the loading flow so the empty view no longer flashes before data finishes loading.
I adjusted hero.tsx for responsive alignment and updated first_run.tsx to separate the loading and empty states.
An isolated, end-to-end agent workflow for your codebase.
Threads keep the task, context, and feedback together from the first prompt to final review.
Connect your repo
Connect a GitHub repository and choose the branch to work from.
Start a thread
Describe the task with the context your agent needs.
Spin up an isolated sandbox
Every task gets its own cloud sandbox and Git worktree, isolated from your original repository.
Let the agent work
The agent reads code, makes changes, and runs tools and commands in that environment.
Review and ship
Review the diff, leave feedback, iterate, and open a pull request when ready.
Run agents in parallel.
Keep every task isolated.
Agents get their own cloud computers to build, test, and complete tasks end to end for you to review.
Bring your own key
Connect the providers you already use with your own API keys. LegionCode orchestrates each run; your provider account and billing stay yours.
Google Gemini
GEMINI_API_KEY
OpenAI
OPENAI_API_KEY
Anthropic Claude
ANTHROPIC_API_KEY
OpenCode
OPENCODE_API_KEY
OpenRouter
OPENROUTER_API_KEY
Cloudflare AI
CLOUDFLARE_API_TOKEN