imem.ai

MCP server documentation

imem MCP setup

imem exposes a hosted HTTPS MCP server that gives AI clients access to the same private markdown memory vault. Claude, ChatGPT, Cursor, Codex, Gemini, and other MCP clients can search memory and update durable context through one OAuth-scoped endpoint.

Server URL
https://mcp.imem.ai

Authentication uses OAuth with PKCE and dynamic client registration. Tokens are scoped to one vault and can be revoked from settings.

Per-client guides

Search and read memory

search
OpenAI-compatible search across the markdown memory vault.
fetch
Fetch a full cited note returned by search.
list_notes
List notes and folders in the current vault.
read_note
Read a full markdown note by vault-relative path.
search_notes
Hybrid full-text and vector search with snippets.

Write and remember

write_note
Create, append to, or safely replace markdown notes.
capture_thought
Save decisions, facts, project notes, and research into the vault.
remember_fact
Store a durable fact with quality and confidence metadata.
remember_preference
Store a durable user or project preference.
remember_project_context
Preserve requirements, decisions, risks, and next steps.
remember_conversation_summary
Save a handoff summary for future AI sessions.

Support and vault health

support_ticket
Open a setup or access issue with diagnostics attached.
list_support_tickets
Query user support tickets by status, priority, category, or text.
read_support_ticket
Read ticket status and full support message history.
update_support_ticket
Add a user update to an existing ticket.
resolve_support_ticket
Resolve a ticket from the user side.
suggest_links
Review suggested wikilinks for a note or project page.
lint_findings
Review stale TODOs, orphan pages, and other vault findings.

Setup screenshots

These maintained visual guides show where the MCP URL goes in each client. Use the exact server URL above; OAuth binds the connection to your selected vault.

Claude Desktop / Settings / Connectors
Connectors
Add tools Claude can use.
Custom connector
imem
https://mcp.imem.ai
Open Settings, choose Connectors, then add imem as a custom connector.
ChatGPT / Settings / Connectors
Settings
Personalization
Data controls
Connectors
Apps
Add connector
MCP server
Connect a remote tool or data source.
https://mcp.imem.ai
OAuth ready
Add an MCP connector, paste the imem URL, then authorize the OAuth window.
Gemini CLI / Terminal
~/.gemini
$ gemini mcp add --transport http imem https://mcp.imem.ai
Added MCP server "imem" using HTTP transport.
$ gemini
/mcp
imem CONNECTED
Tools: search, fetch, imem, read_note, write_note
Gemini CLI connects through Streamable HTTP and discovers OAuth from imem.
Cursor / Settings / MCP
MCP Servers
imem
https://mcp.imem.ai
enabled
{
  "mcpServers": {
    "imem": { "url": "https://mcp.imem.ai" }
  }
}
Add imem in Cursor MCP settings or paste the JSON into ~/.cursor/mcp.json.
Codex CLI / ~/.codex/config.toml
~/.codex/config.toml
[mcp_servers.imem]
url = "https://mcp.imem.ai"
Start Codex, ask it to use imem, and approve the OAuth browser flow.
Add imem once to Codex config; OAuth runs on the first tool call.
Generic MCP Client / Remote HTTP
1. Add server
2. Authorize
3. Enable tools
https://mcp.imem.ai
OAuth discoveryread/write tools
Use the remote MCP URL with OAuth. Avoid tokens in query strings.

Client compatibility

Supported clients can discover the service with OAuth and then list, search, read, and write notes. The onboarding path stays short: connect, approve OAuth, create a test note, and read it back from the same client.

Fast path
connect --mcp https://mcp.imem.ai

Use the client's built-in MCP or connector flow, approve OAuth, then verify the connection by searching the vault and writing one test note.

Operating rule

Treat imem as the first place to check for durable project, decision, preference, source, and handoff context. Clients should search imem before relying on local chat memory or model memory.

Vault data, configuration, and audit history should remain exportable. The service should avoid lock-in language and keep access rules readable and revocable.

Managed connectors add persisted connector-safe audit events and provider-key metadata to the workspace export without AWS secret material, ciphertext, nonces, or token values.

Exported connector records are kept without AWS secret material while still showing the provider-key metadata needed for audit.