Skip to main content
Version: main

GigaAgent overview

GigaAgent is a modular agent framework with a FastAPI backend, a LangGraph execution graph, and a bundled React web UI. The giga_agent dev command starts a local server on port 9090 and serves both the API and the UI.

Main parts​

PartPurpose
giga_agent Python packageCLI command, backend, execution graph, modules, and migrations.
Web UIChat, settings, RAG collections, memory, administration, and files.
Execution graphControls model calls, tool calls, streaming, and callbacks.
ModulesAdd authentication, RAG, memory, search, images, file operations, skills, deep research, and integrations.
ProvidersLanguage models, embeddings, search, image generation, and sandbox runtimes.

What the agent can do​

Useful work requires a configured language model. Depending on user settings, the agent can use chat streaming with clarifying questions, files, RAG, long-term memory, agent skills, deep research, images, code execution, weather, search, and legacy subagents kept for compatibility (landing page, presentation, meme, and more).

External services connect through the connectors catalog: MCP servers, Yandex Mail, Calendar, and Disk, VK, GitHub. Results of some tools render as chat widgets. The agent also runs scheduled tasks with delivery to a Telegram channel.

Start here​

  1. Install and run the repository build: Local quickstart.
  2. Configure a model and send the first message: First chat.
  3. Review feature prerequisites: Capabilities and requirements.
  4. For shared installations, read Shared server.
  5. For implementation details, read Architecture.