Start with context
Bring prior decisions and constraints into a new conversation instead of rebuilding the background every time.
Durable, curated AI memory
Memory that carries the useful parts forward.
Gothel gives the AI tools you trust a shared memory for decisions, preferences, constraints, and project context—without treating every conversation as something worth keeping.
Why it helps
Continuity without the clutter.
Gothel is designed for the small set of things that should survive a chat: what was decided, what matters, and what a future collaborator would otherwise have to rediscover.
Keep memory intentional
Proposals can be inspected before storage, and bulk changes are previewed before anything canonical changes.
Share one source of truth
ChatGPT, Codex, and other compatible clients can use the same tenant-isolated memory through a standard MCP connection.
Core features
Built for careful remembering.
The interface is small on purpose: retrieval is easy, while writes are explicit and auditable.
Search and project context
Retrieve relevant memories, established decisions, architecture, and important constraints.
Proposal-first capture
Screen a candidate for provenance and likely secrets before choosing whether to remember it.
Corrections and history
Correct, archive, or supersede knowledge while retaining attribution and revision history.
Bulk reconciliation
Preview up to 100 candidates, apply the exact saved plan, and resolve ambiguous matches through explicit review.
Optional project scope
Keep personal continuity without a repository, or bind curated project memory to a defined project.
OAuth and tenant isolation
Every memory operation requires authorization; identity and tenant selection never come from model-supplied arguments.
How it works
A clear boundary between connection and memory.
Your client asks
ChatGPT, Codex, or another MCP client calls a focused retrieval, proposal, or memory tool.
Tyrion connects
The public edge verifies OAuth permissions, applies curation policy, and forwards only a verified identity.
Gothel remembers
The private canonical service owns matching, persistence, tenant isolation, revisions, and audit history.
-
1
Confirm your invitation
Ask the person who shared this page to confirm that your identity and ChatGPT connection are approved.
-
2
Enable Developer mode
In ChatGPT, open Settings, choose Security and login, and turn on Developer mode.
-
3
Add the MCP server
Open ChatGPT Plugins, select the plus button, and enter the endpoint above. Choose OAuth when asked.
-
4
Pause if ChatGPT shows new connection details
Send any displayed callback or connection details to the Gothel owner. Continue only after they confirm the connection is registered.
-
5
Sign in and test retrieval
Complete OAuth, start a fresh chat, enable Gothel, and ask it to search memory for a harmless topic. ChatGPT will request confirmation before write actions.
Interface labels can change. See OpenAI’s current developer-mode instructions if your ChatGPT menus differ.
Useful starting prompts
Tell the client how careful to be.
“Search my memory for decisions about authentication before you answer.”
“Review this as a memory candidate. Don’t save it until I confirm.”
“Remember this preference after I approve the exact wording.”
“Preview reconciliation for these memories. Show me the report, but don’t apply it yet.”
Access and privacy
Publicly discoverable. Privately usable.
The health page, protocol discovery, and tool catalog are public so compatible clients can connect. Every memory read, proposal, write, correction, sync, and reconciliation operation requires OAuth authorization.
- Invite-only identity provisioning
- Separate read, write, and project-sync permissions
- Tenant identity derived from verified OAuth claims
- No analytics, trackers, or third-party scripts on this page
- Never use memory to store passwords, API keys, or private tokens