Skip to content

README.md

Latest commit

 

History

History
129 lines (97 loc) · 3.84 KB

README.md

File metadata and controls

129 lines (97 loc) · 3.84 KB

@agent-dispatch/cli

npm CI license

Command-line tools for configuring, validating, and testing AgentDispatch before you connect it to an MCP-capable lead agent.

Install

npm install -g @agent-dispatch/cli

Quickstart

Create a local config:

agentdispatch init \
  --region us-west-2 \
  --runtime-arn arn:aws:bedrock-agentcore:us-west-2:123456789012:runtime/my-runtime \
  --protocol a2a

Validate local configuration and provider requirements:

agentdispatch doctor --config ./agentdispatch.config.json

Run an opt-in AWS live preflight before first AgentCore spawn:

agentdispatch doctor \
  --config ./agentdispatch.config.json \
  --aws-live \
  --runtime research-agent

This resolves AWS credentials through the standard AWS SDK chain and checks that the configured AgentCore runtime is reachable for session-mode dispatch.

Run a smoke-test task:

agentdispatch run \
  --config ./agentdispatch.config.json \
  --instruction "Summarize the current repository and identify follow-up work." \
  --provider aws \
  --account-profile dev-aws \
  --protocol a2a

Pass AgentCore runtime details without hand-writing target.details JSON:

agentdispatch run \
  --config ./agentdispatch.config.json \
  --instruction "Run the background research task in AgentCore." \
  --target-mode session \
  --runtime-arn arn:aws:bedrock-agentcore:us-west-2:123456789012:runtime/my-runtime

For runtime provisioning, use --target-mode runtime with --ecr-image-uri and --execution-role-arn:

agentdispatch run \
  --config ./agentdispatch.config.json \
  --instruction "Run the background research task in a fresh AgentCore runtime." \
  --target-mode runtime \
  --ecr-image-uri 123456789012.dkr.ecr.us-west-2.amazonaws.com/agentdispatch-worker:latest \
  --execution-role-arn arn:aws:iam::123456789012:role/AgentDispatchAgentCoreRole \
  --env AGENT_FRAMEWORK=openclaw \
  --cleanup-after-task

Use --env KEY=value for runtime environment variables such as framework selectors, and add --cleanup-after-task when the runtime should be deleted after completion.

Inspect a task:

agentdispatch status task_...
agentdispatch logs task_...
agentdispatch result task_...

Continue an A2A cloud-agent session from the saved task metadata:

agentdispatch a2a-send task_... \
  --config ./agentdispatch.config.json \
  --text "Continue the investigation and focus on IAM findings."

What the CLI is for

  • Bootstrap agentdispatch.config.json without hand-writing JSON.
  • Verify account profiles, adapter config, runtime mode, and protocol settings.
  • Test spawn_cloud_agent outside a lead-agent environment.
  • Test A2A follow-up against the cloud agent returned by an AgentCore task.
  • Give developers a reproducible command path before wiring OpenClaw, Hermes Agent, Claude Code, or Codex through MCP.

MCP handoff

After doctor passes, point your MCP client at the server:

{
  "mcpServers": {
    "agentdispatch": {
      "command": "npx",
      "args": ["-y", "@agent-dispatch/mcp-server", "--config", "/absolute/path/agentdispatch.config.json"]
    }
  }
}

The lead agent can then call spawn_cloud_agent and receive task polling tools plus A2A cloud-agent metadata.

Development

npm install
npm run typecheck
npm test
npm run build

See the release workflow for npm Trusted Publisher setup, provenance publishing, and upstream package order.