Agent Forge: Parallel AI Agents
Loading...
Skip to content
Agent Forge parallel agent runtime
—

orchestrator

Mission control

Describe an outcome. The lead agent plans, parallel workers run with tools, then everything is merged into one deliverable.

Mission
Parallel agents
Stop run Run agents

No run yet

Start with an example above or write your own mission. Everything the agents build is saved to your workspace files.

  1. Plan
  2. Parallel agents
  3. Synthesis

—

Deliverable

Copy
The run stopped.

virtual filesystem

Workspace

A shared disk your agents can read and write. Everything here persists between sessions.

New file path Create Search files

    No files match. Create one above.

    No file selected
    Save Delete
    File content

    tool registry

    Tools

    Every capability the agents can call, with its JSON schema. Run any tool by hand to see exactly what the agent sees.

    No tools loaded.

    model context protocol

    MCP connectors

    Attach external tool servers. Import a manifest, and their tools join the agent toolbelt instantly.

    Add a connector

    Server name
    Transport
    Endpoint URL
    Tool manifest (JSON)

    Each tool needs a name, description and inputSchema. Tools are exposed to agents as server.tool.

    Add connector

    Connected servers

    0

    No connectors yet. Add one on the left.

    trace log

    Run history

    Every mission with its plan, agent findings and final output.

    Clear history

    No runs recorded yet.

    Runtime settings

    Reasoning model

    Used for planning, every worker and the final synthesis.

    Max tool calls per agent
    Save settings