AI AGENT STORE / SERVICES

MEMORY & CONTEXT · FOR AI AGENTS

Supermemory

Retain project context, retrieve source documents and share memory across authorized agent sessions.

Supermemory provides managed memory and document retrieval through API, SDKs and remote MCP. Namespaces and spaces organize context for a user or project, while separate retrieval tools expose memories, profiles and source documents.

Your integration checklist

What you need from your owner
Create an account, decide which data may be retained and choose a space or namespace. Approve MCP access to the intended spaces, or create an API key in the developer console for application integration.
Authentication
MCP at https://mcp.supermemory.ai/mcp uses browser OAuth and space permissions. API and SDK access uses a secret API key; HTTP examples use Authorization: Bearer.
Your first task
Save a small approved project note in a test space, wait for processing, then retrieve it from a new session and compare the returned memory with its source.
Billing & payment
Plans include monthly credits consumed by usage, with a limited free tier and paid upgrades. Paid auto top-up has configurable purchase limits. Some connectors and deployment options require higher tiers; review the current rate card before bulk ingestion.
Agent payment support
Not verified. Programmatic usage does not establish autonomous checkout support.
Operational limits
Document indexing and memory generation are asynchronous and can finish at different times. Namespace selection must stay aligned with account permissions. Retrieved or inferred memories can be incomplete or outdated; they are not authoritative evidence by themselves.

ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK

Keep memory attributable and test corrections across sessions

Editorial recommendation · Reviewed 2026-11-04

When to choose this service

Choose Supermemory when authorized context must persist across assistants or sessions with managed retrieval. Existing harness memory or a small local file may already meet a single-agent need.

The hosted service maintains shared persistence, indexing and memory processing beyond the agent runtime. Its value is operating that infrastructure at scale, not replacing the agent’s ordinary ability to take notes.

Suggested workflow

  1. Assign a stable namespace to the intended user or project and ingest a small approved source with a stable identifier. Keep other users’ context separate.
  2. Wait for the relevant processing state before testing recall. Use document retrieval for source text, memory search for remembered facts and profile retrieval for ongoing context.
  3. Introduce a correction and verify what a fresh session retrieves. Preserve source references and explicitly remove obsolete entries when the workflow requires it.

Acceptance checks before real work

  • A queued ingestion response does not mean memory is ready.
  • A plausible memory should be checked against its source before a consequential action.
  • Test that the connected client cannot retrieve another space’s restricted content.

If the workflow fails

If recall is empty, check namespace, permissions and processing status before adding duplicates. Inspect source documents for incorrect facts and verify the correction or forget result rather than assuming a successful request changed every derived view.

Provider documentation

Explore services for adjacent needs

Continue building your toolkit