One command.
Complete delivery.
felix v3 run enforces planning, building, and validation in sequence — so nothing lands unless it's been properly tested.
Why a CLI?
Chat-based tools give you suggestions. Felix gives you a delivery process. Felix 3 runs locally, keeps run artifacts in your project, and needs no Felix Cloud account for local specs.
Spec → Plan
Before a single line changes, Felix writes an explicit implementation plan. You can inspect it, reject it, or let it proceed.
Plan → Code
Felix delegates planning, building, and review to your selected harness, such as Claude Code, OpenCode, or Codex. The plan stays the contract.
Code → Validation
Configured validation commands must pass, then Felix runs an independent review against the spec and plan.
Validation → Commit
A clean diff lands in your repo with a structured commit message. The run artifacts — plan, patches, logs — are there to audit.
Runs on your machine
No cloud execution. Felix uses your local environment, your tools, your secrets. Nothing is proxied.
Swap your agent freely
Profiles let you run different LLMs per project or requirement. Felix doesn't lock you into a provider.
State lives in files
Progress is stored as files in your repo. Restart anytime, resume from where you left off, or roll back with git.
Scales to a swarm
Run one agent locally or register multiple across machines. The sync server coordinates without touching your code.
Sync is optional.
Operate fully local with durable file-based artifacts, or enable sync for team visibility. Either way, your code never leaves your machine.
Full reference in Docs
Every command, flag, config option, and agent profile setting is documented. The docs tab has everything you need to go from install to first run.