Developer docs
Learn the concepts, then build with the API.
These docs explain how Copepod works and describe the service APIs used by agents and the portal.
Overview
What Copepod does
Consistent memory updates
Memory writes are queued and consolidated, with tenant and agent context applied to each request.
Durable contexts
Build bounded private context snapshots from authorised evidence, with citations and revision history.
Workspace security
Tenant scoping, audit events, rate limits and step-up controls protect workspace operations.
Structured knowledge
Define schemas, store partial records and enrich missing fields over time.
Agent identity
Agent-bound API keys identify the caller and workspace, so clients do not need to send identity fields.
Libraries & SDKs
Use the REST API from curl, Python or any HTTP client. The Python SDK is in private beta; Node and Go clients are planned.
MCP Server
MCP clients can store, search and link memory, then build durable context for agent workflows.
Controlled sharing
Keep working memory private to an agent, then publish reviewed knowledge to the shared workspace.
Concepts
Core product terms
Tenant
A tenant is a Copepod workspace. The server derives the tenant from the authenticated session or API key and keeps its data separate from other workspaces.
Memory
A memory is a versioned record with content, tags, visibility, source history, conflict state and enrichment metadata. It can be shared with the workspace or private to an agent.
Knowledge structure
A schema that defines the fields for a knowledge object and the enrichment rules for each field.
API key
A workload API key is bound to an active agent and limited by scope. The key supplies the agent and tenant identity and can be rotated or revoked.
MCP Server
A Streamable HTTP Model Context Protocol endpoint for compatible clients. Its tools cover memory operations, search, durable contexts, memory links, grounded briefs and service status.
Durable context
A private, versioned context capsule built from authorised evidence. It has a fixed token budget, citations and a snapshot version. The builder extracts evidence without asking a generative model to compose the result.
API reference structure
Each service has its own reference page, so you can find the endpoints, request fields and responses needed for your integration.
Quick start
curl -H "X-API-Key: <your-api-key>" \ "https://api.copepod.ai/memories/search?q=context&k=10"
Use an agent API key for direct integrations. The tenant and agent identity come from the key, so you should not send a tenant or agent ID.