Coming from the AgentCore starter toolkit
llms.txt lists every page for a coding agent.
Show the prompt
Read https://intentius.io/loomster/guides/coming-from-agentcore-starter-toolkit/ and, against the person's local Loom (`just local-up`), show the Loom screen or API call for each toolkit command they use.
If they name an AgentCore runtime they deployed with the toolkit, show how to register it in Loom by ARN; register it only if they ask in this session.
Never run `chant approve` or `just grant`, never deploy to or delete from a real AWS account unless the person asked for that in this session, never register an identity provider in Loom, and never push or edit the `chant/lifecycle` branch.Result
Section titled “Result”An agent you built with the toolkit, or one Loom builds for you, governed in Loom: in its catalog, behind its groups and approval policies, with its costs, evaluations and skills.
Loom’s additions
Section titled “Loom’s additions”The starter toolkit gets one agent onto AgentCore Runtime from your laptop. Loom is the platform around many agents: users and groups from Cognito, a catalog, IAM roles registered by a security admin, approval policies on tool calls, evaluations, skills from an Agent Registry, MCP servers and A2A agents, and cost and activity dashboards. loomster deploys Loom itself.
The first command
Section titled “The first command”just local-up # Loom on your laptop, no AWS accountThen open http://localhost:8080 (Run Loom locally).
Agents deploy on Floci too; invoking one
there returns a canned reply.
Commands
Section titled “Commands”| To do this | Starter toolkit | Loom |
|---|---|---|
| Describe an agent | agentcore configure --entrypoint agent.py | Agents > new agent, guided setup (Runtime, Identity and models, Access, Tools and memory, Lifecycle and tags), or import a JSON manifest |
| Deploy it | agentcore launch | the end of the guided setup, or POST /api/agents with source: "deploy" |
| Bring an existing runtime | register it by ARN: the new-agent screen’s “register an existing runtime”, or POST /api/agents with {"source": "register", "arn": "<runtime ARN>"} | |
| Invoke | agentcore invoke '{"prompt": "..."}' | the agent’s Invoke tab, or POST /api/agents/<id>/invoke |
| Check status | agentcore status | the Agents screen, or GET /api/agents/<id>/status |
| Remove | agentcore destroy | delete the agent in Loom (DELETE /api/agents/<id>) |
| Execution role | created for you | registered by an admin under Security > IAM roles; loomster creates it (Create an execution role per agent) |
| Memory | configured with the agent | the Memory screen |
Your own agents
Section titled “Your own agents”- Loom’s deploy path builds agents from configuration on its own harnesses, Strands by default, with Google’s ADK as an option. You choose the model, tools, memory and access; no code of yours is deployed.
- An agent with your own code, deployed with the toolkit or any other way, is a runtime ARN. Register it in Loom to put it in the catalog and invoke it through Loom.
- The assistant loomster deploys in
loom-agentsis a chant-managed runtime from Loom’s Strands harness on AgentCore Runtime V2, as Loom’s own deploys are.
The roles a registered runtime runs with are yours; the Loom role picker lists only roles registered under Security > IAM roles.
Behavior differences
Section titled “Behavior differences”- Loom never creates IAM. The toolkit makes an execution role for you; in Loom an admin registers one and tags it into the group allowed to use it.
- A user sees only agents whose
loom:grouptag matches one of their groups. - Model calls go through
bedrock-runtime, and for some modelsbedrock-mantle; both need egress on a private network.
- Browse Loom on your laptop.
- Screens for what each screen does.