Kodo

Building while you sleep.

🦉 kodo

Autonomous multi-agent coding that runs overnight using your installed coding agents. An orchestrator delegates implementation and review across work cycles, with resumable logs and checkpoints.

Verification is currently directed by the orchestrator: the default completion tool asks it to consult testers, but does not enforce an independent verification gate. Check the run's test and review evidence before relying on its success status.

SWE-bench Verified: Kodo 57% vs Cursor 46%

On a 100-task head-to-head using the same underlying model (Cursor composer-1.5), adding Kodo's orchestration layer solves 24% more real-world GitHub issues. Same model, same prompt, same conditions — the difference is orchestration. Full methodology and interactive results →

Overview

See detailed mode diagrams for the full pipeline of each mode.

🎬 How it works in practice

Real run from blackopt — building an auto-solving meta-optimizer with 4 new algorithms, adaptive scheduling, and 73 tests. 3 hours unattended, 2 cycles, succeeded.

The architect verifier caught 9 rounds of bugs that the worker agent was blind to — resource leaks, class variable contamination, state mutation — each subtler than the last. A single Claude Code session would likely have shipped with several of these.

🦉 When to use kodo

You have a Claude Code Max subscription. You can't use it while you sleep.

kodo lets you set a goal and leave a team working through multiple cycles. Choose an API orchestrator such as Gemini Flash or Claude, or an installed CLI backend. The orchestrator delegates implementation, testing, and review to the team.

🌙 Overnight runsSet a goal, leave it running for hours. Cycles checkpoint progress automatically.
🔍 Review rolesArchitect and tester…