Remember why, not just what
Most memory systems store what was said. Fabric stores what was decided, who owed what to whom, what evidence backed it, and what happened next — as linked, queryable entries.
- 7 entry types
- 5 endpoints
- Webhook on every write
- Optional linked semantic memory
The question that has no good answer today
Six months into a project, somebody asks why the team chose this database. The decision is real, but the reasoning is scattered across a chat thread, a document nobody can find and one person's memory.
Seven kinds of entry, one shape
A Fabric entry is not free text with a label. Each type carries the fields that make it answerable later — an assignee and a status on a task, an evidence list on a decision, a verified flag on a review.
Entries are scoped by project and customer, so a question about one engagement never returns another's history. Every write can also mint a linked semantic memory, which means Fabric entries surface in ordinary retrieval alongside everything else rather than sitting in a silo.
Each write fires a fabric.{type}.created
webhook, so a project tracker or a notification bot can follow the
work without polling.
fabric_entries
A decision entry as stored
- entry_type
- decision
- title
- Move the operator dashboard off SSR
- project
- memory-os-dashboard
- decided_by
- D. Okafor
- evidence
- fe_18b3 · m_77e1
- verified
- true
- created_at
- 2026-04-09T14:22:11Z
POST /v1/fabric/write → fabric.decision.created
What you ask it
Five endpoints, each answering a question a team actually asks out loud.
| Endpoint | What it answers |
|---|---|
POST /v1/fabric/write | Record a decision, task, review, outcome, procedure, handoff or note, with assignee, project, evidence links and artifact paths. |
POST /v1/fabric/recall | Retrieve the Fabric entries relevant to a question, scoped by project and customer. |
GET /v1/fabric/pending | The open-task list for a user, agent or customer. |
GET /v1/fabric/timeline | The ordered history of a project. |
POST /v1/fabric/why | Return the decision-and-evidence chain behind a conclusion. |
Your agent can write to it directly
Four of the eighteen MCP tools are Fabric tools, so a coding agent can record the decision it just made, check what is still open, and ask why something was decided — without you building an integration.
pcnaid_fabric_write— record a decision or taskpcnaid_fabric_recall— retrieve relevant entriespcnaid_fabric_pending— list open workpcnaid_fabric_why— return the evidence chain
{ "query": "why did we drop SSR?" }
// → the chain, not a paraphrase
decision Move the dashboard off SSR
evidence Build timing comparison
outcome Cold build 4m10s → 38s
memory m_77e1 (linked)
// scoped to project + customer
Give your agents a memory you can audit
Run the whole system on your own hardware under the MIT licence, or ask us about hosted access. Both start from the same place.