Repolode
Submit
keta1930

keta1930/agent-graph

A multi-agent system built on Context Engineering, with a visual editor for designing, testing and deploying agent workflows.

View on GitHub
Stars
234
Forks
46
Open issues
5
Last commit
11 days ago

What is it?

Agent Graph is a multi-agent system for building AI applications from several cooperating agents. It combines sub-agents, long-term memory, MCP tool integration and agent-based workflows, and applies context engineering practices to the way agents receive and manage information.

Workflows are designed in a visual graph editor with linear, parallel, conditional and nested structures. Graphs can be reused as subgraphs, agents can hand off tasks to the next node dynamically, and graphs can run on a cron schedule with execution history. Models are configured through OpenAI-compatible APIs, with support for both LLM and VLM models.

The project is deployed with Docker Compose for supporting services such as MongoDB and MinIO, with the FastAPI backend run via uv or pip. It includes team management with invitation codes and role-based permissions, plus a prompt library and project-based conversation organization.

Key features

  • 01Visual drag-and-drop graph editor for linear, parallel, conditional and nested workflows.
  • 02Subgraph nesting and dynamic handoffs between nodes.
  • 03Scheduled Graph execution with cron expressions and run history.
  • 04MCP protocol integration for connecting external tools and data sources.
  • 05Short-term and long-term memory for agents, with a prompt template center.
  • 06Team invitations and role-based permissions.

Use cases

  • Building multi-step agent pipelines where several specialized agents hand work to each other.
  • Scheduling recurring agent jobs that run a workflow on a cron schedule.
  • Giving agents access to databases, APIs or file systems through MCP servers.
  • Running a shared agent platform for a small team with managed accounts and permissions.

Topics

#agent#multi-agent#workflow#mcp#llm#context-engineering#memory#agentic-framework#agentic-workflow#graph

FAQ

What is agent-graph?
A multi-agent system built on Context Engineering, with a visual editor for designing, testing and deploying agent workflows. Agent Graph is a multi-agent system for building AI applications from several cooperating agents. It combines sub-agents, long-term memory, MCP tool integration and agent-based workflows, and applies context engineering practices to the way agents receive and manage information.
Is agent-graph free to use? What license does it have?
agent-graph is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is agent-graph still maintained?
Actively maintained. The last commit to agent-graph was on Oct 11, 2026.
What can I use agent-graph for?
Building multi-step agent pipelines where several specialized agents hand work to each other. Scheduling recurring agent jobs that run a workflow on a cron schedule. Giving agents access to databases, APIs or file systems through MCP servers. Running a shared agent platform for a small team with managed accounts and permissions.
Where can I get agent-graph?
The source code and installation instructions are on GitHub: https://github.com/keta1930/agent-graph

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects