Skip to main content
Glama
sabbour

Agentweaver MCP Server

by sabbour
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Hybrid

Agentweaver

⚠️ Alpha software. Agentweaver is under active development. Expect breaking changes and incomplete features. Do not use it in production.

Run teams of AI agents on your own infrastructure.

Describe the work you want done. Agentweaver can generate the agent roles, skills, and workflow for that work, then run the team in isolated environments on infrastructure you control.

Workflows turn probabilistic agent work into a governed path toward a defined outcome, with the gates and approvals you set. Start and supervise work in the Agentweaver interface or through MCP from an assistant, editor, or CLI.

Visual model

System architecture

Deployment and component view showing web and MCP clients entering the Agentweaver API, API and worker orchestration authority, isolated AgentHost execution, durable PostgreSQL and Azure Files state, and separate identity, repository, and model-provider dependencies.

Structured source · Editable draw.io

📖 Read the documentation or browse the source in docs/index.md.

Related MCP server: harness_to_mcp

Why Agentweaver

  • Generate the team and process. Start from a description or reusable blueprint to define roles, skills, and the workflow the agents follow.

  • Set review gates. Confirm the intended outcome before work starts and approve the assembled result before it merges.

  • Inspect every run. Watch agent events and topology live. Runs keep their work in isolated git worktrees.

  • Use your preferred surface. Operate the same projects, teams, workflows, and runs from the web interface or an MCP client.

  • Carry context forward. Team memory and decisions give later runs prior constraints and learnings.

What it includes

Capability

What it does

Projects

Binds a repository and AI configuration to teams, runs, and memory.

Teams and blueprints

Defines reusable roles, policies, and agent teams.

Workflows

Runs configurable multi-role flows for delivery, research, content, operations, and evaluation.

Board and runs

Organizes work and shows active orchestration from plan to review.

Review

Uses an RAI check and human approval before a result merges.

MCP server

Lets MCP clients work with Agentweaver through authenticated tools.

Quick start

Install Git, Node.js, and the .NET 10 SDK. On Windows, install WSL2 and bubblewrap before you run local agent work.

git clone https://github.com/sabbour/agentweaver.git
cd agentweaver
npm run setup
npm run dev

npm run setup prepares the local application. npm run dev starts the API at http://localhost:5000 and the web UI at http://localhost:5173.

Before you sign in, configure local authentication and model access. The Getting started guide has platform setup, authentication, and first-run instructions.

Learn more

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityActive
ResponsivenessResponsive

Resources

Related MCP Connectors

Related MCP Servers