Tenka Memory — connector documentation
Tenka Memory is a private, user-owned memory for AI assistants. From inside Claude or
ChatGPT, a user can save the current conversation into their own Tenka Memory and recall
it later — in a new chat, or in the other assistant. This page is the public reference
for users and for IT administrators evaluating the connector.
- Connector (MCP) endpoint:
https://dashboard.tenka.cloud/mcp/memory - Provider: Tenka AB (Sweden, EU)
- Support: jon@tenka.se
- Privacy policy: https://tenka.cloud/privacy · Terms: https://tenka.cloud/terms
What it does
- Save a conversation into a project the user chooses (or a default "Saved chats" project).
- Recall previously saved chats and notes by search.
- Organize saves into projects and get a compact project summary to reload context.
- Export a project (or a single note) to a styled Word
.docxvia a short-lived link.
What it cannot do
- No background capture. Saving is model-mediated and on-request: the assistant can
only pass along a conversation the user explicitly asks it to save, in a chat where the
connector is enabled. It cannot read the user's other chats or past conversations.
- No deletion through the connector. The AI surface can add to and read a user's
memory but exposes no delete operation. Permanent deletion happens only in the user's
own dashboard at tenka.cloud.
- No access to the wider platform. Connector tokens are scoped to Memory only — they
cannot reach servers, billing, or account settings, even though Tenka Cloud offers those
to signed-in users elsewhere.
Tools exposed
| Tool | Access | What it does |
|---|---|---|
list_projects | read-only | List the user's Memory projects (id, name, role). |
save_conversation | write (non-destructive) | Save a chat transcript as note(s) in a chosen project. Confirms before writing to an existing non-default project. |
get_context | read-only | Return a compact summary of one project to reload context. |
search_notes | read-only | Search notes within a project by text. |
get_docx_link | read-only | Return a short-lived link to download a project or note as Word .docx. |
Every tool carries a title and read-only / non-destructive annotations. There are no
destructive (delete) tools on this surface.
Authentication & access control
- OAuth 2.1 with PKCE. The connector self-registers via Dynamic Client Registration
(RFC 7591); there is no shared client secret to distribute.
- The user signs in with their own Tenka account and approves access on a consent screen
that states the connector is limited to Memory.
- Tokens are Memory-scoped. Access to any project is checked per request against the
user's membership (owner / editor / viewer); the connector can never cross between
different users' or tenants' data.
- Transport is Streamable HTTP over HTTPS.
Data handling
- Data a user saves is stored in their Tenka Cloud account, hosted in the EU.
- Writes are rate-limited and scanned for accidentally-pasted secrets (a non-blocking
warning), and every save is recorded in the user's audit log.
- Users can review, edit, export, and permanently delete their memory from the
dashboard at any time. See the privacy policy for retention and GDPR rights.
How to connect
Claude (Claude.ai / Desktop / Code): Settings → Connectors → Add custom connector →
enter https://dashboard.tenka.cloud/mcp/memory → sign in with Tenka. Then, in any chat
with the connector enabled: "Save this conversation to Tenka Memory."
ChatGPT: add the same MCP endpoint as a connector, or importhttps://dashboard.tenka.cloud/openapi-memory.json as an Action in a custom GPT (OAuth).
Once Tenka Memory is listed in the Claude and ChatGPT app directories, it can be added in
one click without pasting a URL.
Support
Questions, access issues, or data requests: jon@tenka.se. We aim to respond within
two business days.