Evaluate for your org
llms.txt lists every page for a coding agent.
- Three tiers from one source: light for evaluation, production and production-ha for real use.
- Many Looms side by side, one instance per account or boundary, optionally sharing one Cognito pool.
- Your VPC, KMS key, certificate, zone, repositories, IAM roles, database and Cognito pool through parameters, without forking.
- Backup, restore drills, gated restore, upgrade, rotation, drift detection and teardown as Ops.
- Generated CI for GitHub, GitLab and Forgejo.
Differs
Section titled “Differs”- Loom’s isolation between teams is logical (groups and tags in one database). A hard boundary is a separate instance.
- loomster deploys one account per instance. Several accounts means several instances.
- A deployed tier starts with no users. Sign in to the web app once you add the first admin.
- Agent runtime, model and Evaluations costs are usage-based and come on top of the infrastructure.
First step
Section titled “First step”Deploy the light tier to AWS on a scratch account, then read Validation.
Validation: what has run on each tier and target, the recorded claims, and what nothing has shown.
Then read
Section titled “Then read”Tasks: Go to production, Run a second deployment, Share a Cognito pool, Estimate cost.
Explanation: Tiers and instances, Provision, reference or omit, Choosing chant, Stateful resources.
Reference: Tiers and targets, Cost, Security, Adoption seams.