Introduction
Multi-agent coding in the terminal.
Kolega Code is for developers who already know the terminal-agent workflow and want more control when the work gets wide: Gigacode fan-out, observable sub-agents, reusable custom agents, local sessions and credentials, per-role model routing, web search, and browser automation.
Point it at a project and use it like a normal coding agent for focused work. When a task needs breadth — auditing many packages, migrating a large surface area, running competing reviews, or implementing independent chunks — Kolega Code can split that work across specialized agents and synthesize the results.
Built for fan-out work
Section titled “Built for fan-out work”Gigacode is Kolega Code’s multi-agent workflow layer. When enabled, the agent can write and run small workflows that launch many sub-agents in parallel, pipe work through phases, loop over sets of files or modules, and collect the results into a final answer.
Use it for work such as:
- Auditing every package, module, or service area in parallel.
- Running migration checks across many directories.
- Assigning independent implementation tasks after a plan is agreed.
- Running several review agents against the same change and synthesizing disagreements.
- Pairing browser or investigation agents with code agents when external research and repo inspection both matter.
Workflow runs are visible in the TUI, produce saved artifacts, and can be resumed instead of restarted when intentionally continuing an interrupted run.
Feature map
Section titled “Feature map”- Gigacode orchestration: parallel phases, pipelines, loops, judge panels, synthesis, saved artifacts, and resume support.
- Specialized sub-agents: planning, building/coder, investigation, general, and browser agents with live activity tracking.
- Custom agents: reusable Markdown-defined specialists with focused prompts and optional mode, tool, model, and iteration limits.
- Repo and terminal tools: read/search/edit files, create files, inspect diffs/session changes, and run shell commands with streamed output.
- Web search and browsing: DuckDuckGo works by default with no key; Firecrawl, Tavily, and SearXNG are configurable backends. Kolega Code can fetch URLs directly or drive a Playwright browser for interactive sites.
- Provider and model routing: choose supported providers/models, configure thinking effort, split long-context/fast/thinking roles, and override models per agent role.
- Local-first state: sessions, settings, project permissions, OAuth tokens, and API-key settings stay on your machine.
- Plan/build workflow: use read-only Plan mode for investigation and a reviewable task list, then Build mode to edit and test.
- Interactive or scriptable: use the TUI for day-to-day work or
kolega-code askwith JSON output for automation. - Extensibility: package reusable workflows as Agent Skills, adjust prompts with project overrides, and run lifecycle hooks.
Two ways to use it
Section titled “Two ways to use it”| Mode | Command | Best for |
|---|---|---|
| Interactive TUI | kolega-code . |
Day-to-day development, exploration, orchestration |
| One-shot | kolega-code ask "…" |
Scripting, automation, quick questions, CI |
There are also helper commands for managing sessions and checking your configuration.
Supported providers
Section titled “Supported providers”Kolega Code supports Anthropic, OpenAI API, OpenAI via ChatGPT subscription sign-in, Google, Groq, Together.ai, Fireworks.ai, xAI/Grok, DashScope/Qwen, OpenRouter, Moonshot/Kimi, DeepSeek, Z.AI/GLM Coding Plan, Kimi Coding Plan, Ollama Cloud, and local Llama.
Provider breadth is only part of the control surface: Kolega Code can assign different models to the long-context, fast, and thinking roles, and can also override models for planning, building, investigation, general, and browser agents. See Providers & Models.
What you need
Section titled “What you need”- Python 3.11+
- An API key, ChatGPT sign-in, or local model for at least one supported model provider
- A terminal that supports a modern TUI (most do)
Install
Section titled “Install”Install with the script:
curl -fsSL https://kolega.dev/install-kolega-code.sh | shOr with uv:
uv tool install kolega-codeVerify the install and start a session in your project:
kolega-code --versionkolega-code .See Installation for requirements, other install methods, and upgrade/uninstall commands.
Next steps
Section titled “Next steps”Once you’re installed and have a provider configured, head to the Quick Start to run your first session and try a Gigacode-ready workflow.