remember
Upload files, paste text, or connect a source. trellismem extracts the people, things and relationships and adds them to the graph.
Memory for teams and agents
Shared memory for your team and its AI agents.
trellismem turns your documents, tickets, chats and code into one knowledge graph. People ask it in the app. Agents ask it over MCP or the API. Everyone gets the same answer.
How it works
Four verbs, one graph.
Everything trellismem does is one of four operations on the same memory. The app, the API and every connected agent use the same four.
recall
Ask in plain language. Answers come from the graph and the passages behind it, limited to the datasets you are allowed to read.
improve
Conversations and agent sessions feed back into memory, so the next answer starts from what the last one learned.
forget
Remove a document or a whole dataset, and its facts leave the graph with it.
Sources
Memory that fills itself.
Connect the tools your team already works in. You choose the channels, projects or folders. trellismem reads them and keeps memory in step on every sync.
- Slack
- GitHub
- Linear
- Notion
- Google Drive
- Gmail
- Google Calendar
- Confluence
- Jira
- HubSpot
- Salesforce
- Intercom
- Stripe
- File uploads
- Web pages
Agents
One memory behind every agent.
Give each agent an API key and it recalls from the same graph your team uses, and writes what it learns back. The app generates the setup for each tool with your address and key filled in.
- Claude Code
- Codex
- Claude Desktop
- Cursor
- VS Code
- Gemini CLI
- Cline
- Hermes Agent
- OpenClaw
- n8n
- Dify
- REST API
- MCP
curl -X POST https://api.trellismem.com/api/v1/recall \
-H "X-Api-Key: $TRELLISMEM_API_KEY" \
-H "Content-Type: application/json" \
-d '{"query": "Who owns the Acme renewal?"}'Recall over HTTP. The same call works from any language.
Built for teams
Shared where you want it, private where you need it.
Access per dataset
Each dataset has its own read, write, delete and share permissions. A question only searches the datasets the person or agent asking is allowed to read.
Isolated storage
Every dataset gets its own graph and its own vector space, so one team's memory never leaks into another team's answers.
Passkeys and two-step sign-in
People sign in with a passkey or add an authenticator app as a second step. Agents use API keys you can revoke one by one.
Invite-only workspaces
Close public registration and add teammates from the Members page, so nobody outside your team can create an account.
See the graph
Browse the knowledge graph in the app and see how people, projects and documents connect, grouped by the categories you gave your uploads.
Find the gaps
Memory gap analysis lists the topics in a dataset and the questions it answers worst, so you know what to add next.
Give your team and its agents one memory.
Connect a source, ask a question, and point your first agent at it.