Coding agents

hanzo runs a coding agent against the repository you are standing in. Hanzo Dev is the default. Every other agent is a flag, and naming one gets you that agent or a clear failure.

curl -fsSL hanzo.sh | sh       # install
hanzo                          # sign in, pick your agent, start it
hanzo --claude                 # Claude Code
hanzo --codex                  # Codex
hanzo --gemini                 # Gemini CLI; also --qwen, --opencode, --kilo, --crush, --goose, --aider,
                               # --hermes, --pi, --copilot, --amp, --grok, --cline, --continue, --kimi,
                               # --omp, --letta, --kode, --cursor, --agy
hanzo --claude "fix the failing test in api/users"   # one quoted task runs headless

hanzo dev is the same launcher, and hanzo code is hanzo dev. Each agent keeps its own login and plan. The Hanzo toolset is attached for the run, so the agent can hand work to Enso, Kai and Hanzo's search tools. --cloud moves the agent's own model calls onto Hanzo too. When the session ends, the summary names every hand-off and the model that took it.

Coding tools has the full table of agents, sign-up and plans, what each flag does, and how to read the summary.

Code Intelligence API

hanzo code search, hanzo code context, hanzo code file and hanzo code index are a different thing that shares the word: they are the code intelligence API, search and symbols across your indexed repositories, for you and for agents. They take no agent and run none. The rule is the word after code: an operation of the code index runs it, and anything else starts a session.

The Agent Ecosystem: Dev, Gen & Kata

Hanzo provides three complementary tools for software engineering:

Hanzo Kata: Closed-Loop Test Repair & CI/CD Triage

Kata is an autonomous agent loop engineered specifically for continuous test repair and CI/CD triage. Rather than generating unconstrained prose, Kata uses Kai typed decisions to evaluate failure logs and Enso routing to synthesize AST fixes that pass test suites.

# Run tests and let Kata triage and repair failures
kata run --suite tests/unit

Kata Workflow

1. Detect: Captures failure traces from pytest, cargo test, vitest, or go test.
2. Triage: Asks Kai for a discrete decision (choice: retry, patch, rollback, approve).
3. Synthesize: Routes code edit generation via Enso to the optimal model.
4. Verify: Re-runs the test suite in an isolated sandbox until 100% green verification.

Open Source Repositories & SDK Examples

Every Hanzo agent and client is open source with runnable examples in all tier-one languages:

Hanzo Kata (hanzoai/kata)

• GitHub Repository
• Python Autonomous Loop (examples/python/kata_loop.py)
• TypeScript Autonomous Loop (examples/typescript/kata_loop.ts)
• Go Client (examples/go/main.go)
• Rust Client (examples/rust/src/main.rs)
• Modular Rust Recipes Crate (kata-recipes)

Hanzo Gen (hanzoai/gen)

• GitHub Repository
• Python Assistant & Tool Dispatch (examples/python/gen_tool_calling.py)
• TypeScript Client (examples/typescript/client.ts)
• Go Client (examples/go/main.go)
• Rust Client (examples/rust/src/main.rs)

Hanzo Dev (hanzoai/dev)

• GitHub Repository
• Terminal Installation (curl -fsSL https://hanzo.sh | sh)
• npm Package (@hanzo/dev)