- Home
- Skills
- Workflow Automation
- Marrow Memory: Long-Term Memory for AI Agents
Works with the AI Agents you already use
Marrow Memory: Long-Term Memory for AI Agents
Gives one agent durable memory on the Marrow MCP server: sources, decisions, deadlines and open questions that survive between sessions.
Free
Marrow Memory: Long-Term Memory for AI Agents
Example session with this skill installed
Pick up where we left off on the conference report. Remind me what's still open, and record the keynote summary as approved.
- Read your context and instructions
- Compiled the marrow memory
Summary
Loaded the saved session state. Recorded the keynote summary as approved. Two items remain open.
Recorded this session
| Item | Status | Where |
|---|---|---|
| Keynote summary | Approved (you confirmed) | reports/keynote.md, journal entry added |
Still open (not recorded)
- Panel 2 summary: draft discussed, not confirmed
- Speaker A's slides and the attendee notes disagree on the attendance figure. Both sources are recorded, and I treat the slides as current.
Which do you want to use?
Time-bound
- Submission deadline is in 3 days. The programme dates were last checked 9 days ago and are marked for re-verification.
Files
- Changed:
reports/keynote.md,journal/2026.md,state/session.md - Not touched:
sources/,reports/panel2.md
Next step
Confirm the attendance figure, then I can finalise Panel 2.
Connects securely to your tools. The creator never sees your data.
About this skill
What it does
Most agents start every session blank. This skill teaches an agent to use the Marrow MCP server as long-term memory, so research, writing and planning projects can pick up exactly where they stopped.
It covers one agent working alone, with no code and no multi-agent pipeline.
What the agent will do
- Load the saved session state and next step at the start of every session
- Keep discussed and recorded apart: nothing becomes final until you explicitly confirm it
- Store every fact with its source and date, and label inferences as inferences
- Record conflicts between sources instead of silently picking one
- Flag deadlines and facts whose refresh date has passed
- Treat text you paste in as unverified source material until it is checked
- Keep history append-only and the session state short and current
- End each session by saying what changed and what did not
Requirements
This skill needs the Marrow MCP server connected to your agent. It does not install the server. Source and setup guide: https://github.com/desikai-lab/Marrow
Good for
Research projects, report preparation, planning, notes and deadline tracking, anything where "what did we decide, and based on what?" matters.
Not for
Multi-role coding pipelines with approval gates. Use the companion pipeline skill for those.
Requires the Marrow MCP server, which is a separate install and is not bundled.
Designed for one agent working alone; not for multi-role coding pipelines.
Tool reference written against one server version; later versions may differ.
Tested so far with a limited number of agents.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
Permissions
The skill itself has no direct file, shell or network access. All reads and writes go through the tools of the Marrow MCP server, which the user connects separately.