Datasets
Group memory by project, team or customer. Each dataset is searched, shared and deleted on its own.
Features
Everything your team knows, in one place agents can reach.
trellismem is a memory layer in the cloud: it ingests your team's knowledge, keeps it current, and answers questions from people and AI agents alike. There is nothing to install.
Memory
A knowledge graph, not a pile of chunks.
trellismem reads what you give it and stores what it means: who, what, when, and how they relate. Answers are assembled from those facts and the passages behind them.
Categories
Tag uploads with a category and the graph keeps their entities together, so a question can be scoped to one topic.
Search that picks its own strategy
Ask a question and trellismem chooses how to look: across the graph, in the passages, by keyword, or by date.
Graph models
Describe the entities and relationships your domain cares about, and extraction fills your model instead of a generic one.
Knowledge graph view
Open the graph in the app and see what is connected to what.
Memory gap analysis
See the topics a dataset covers and the questions it answers worst, weakest first.
Sources
Connect the tools your team already uses.
Pick the channels, projects, folders or record types to include. trellismem reads them and picks up changes on every sync. Files, web pages and your own code can add knowledge too.
Slack
Channels and threads
GitHub
Repository code and docs
Linear
Issues, comments and projects
Notion
Pages and databases
Google Drive
Folders and shared drives
Gmail
The labels you select
Google Calendar
Events in the calendars you select
Confluence
Spaces, pages and comments
Jira
Issues and comments
HubSpot
Contacts, companies, deals and tickets
Salesforce
Accounts, contacts, opportunities and cases
Intercom
Conversations, contacts and help articles
Stripe
Customers, invoices and subscriptions
File uploads
PDFs, documents, images and audio
Web pages
Any address you give it
API
Text and files sent from your own code
Agents
Every agent on the same memory.
Each tool below has a guided setup in the app, with your address and API key filled in. Anything else can use the REST API or the MCP server directly.
- Claude Code
- Codex
- Claude Desktop
- Cursor
- VS Code
- Gemini CLI
- Cline
- Hermes Agent
- OpenClaw
- n8n
- Dify
- REST API
- MCP
Learning
Memory that gets better with use.
What people and agents ask is itself knowledge. trellismem keeps it.
Sessions
Conversations keep their own short-term memory, so a follow-up question knows what came before.
Self-improvement
Session answers, agent traces and feedback are folded back into the permanent graph.
Skills
Store your team's playbooks as skills. Agents find the right one for a task and load it on demand.
Activity and analytics
See what was remembered and recalled, and how memory is being used over time.
Access and security
Your team's memory stays your team's.
Permissions are enforced where the data is stored, not only in the interface.
Per-dataset permissions
Grant read, write, delete and share on each dataset to the people who need it.
Isolated storage
Each dataset has its own graph and its own vector space.
Passkeys and two-step sign-in
Sign in with a passkey, or add an authenticator app as a second step, with ten single-use recovery codes.
API keys
Agents and automations use their own keys. Create one per agent and revoke it without touching the others.
Members and invites
Create accounts for teammates from the Members page. There is no public sign-up.
Encrypted connector credentials
Tokens for connected sources are encrypted before they are stored.
Give your team and its agents one memory.
Three steps take you from an empty workspace to an agent that knows your team.