Are you an LLM? Read llms.txt for a summary of the docs, or llms-full.txt for the full context.
Skip to content

Amodal Cloud

Amodal runs agents defined by Markdown, JSON, and TypeScript files. You provide instructions, knowledge, API connections, and tools. Amodal hosts the runtime and provides source editing, deployments, conversation history, evals, logs, and usage tracking.

Source can live in a connected Git repository or in Amodal-hosted snapshots. An agent project starts with amodal.json and an amodal/ directory.

Build an agent

  1. Run locally to try an agent on your machine, or create one in Amodal.
  2. Follow the support-agent tutorial for a complete example.
  3. Deploy and test through the Platform API to automate the workflow.

Define its behavior

ContentPurpose
AgentsInstructions and access to tools, connections, skills, and stores
ConnectionsAPI configuration, authentication, and endpoint policies
SkillsProcedures the agent can load when relevant
KnowledgeReference documents and searchable corpora
ToolsCustom actions implemented in TypeScript or JSON
StoresStructured data the agent can read and write within its permissions

See Project Structure for the file layout.

Deploy, test, and embed

  • Evals: check behavior before deploying changes.
  • Providers: choose models and configure provider access.
  • Authentication: control access to the hosted app and runtime.
  • React SDK: embed chat and agent data in your application.
  • Runtime Apps: deploy a custom frontend at the agent's URL.
  • Runtime Server: call chat endpoints and consume streaming responses.