How does a team share AI memory?

A Basic Memory team workspace gives everyone, and their AI agents, one shared knowledge graph with owner, editor, and viewer roles, version history, and audit logs. Working sessions, plans, and shipping receipts live in one searchable place instead of scattered across chats.

Can AI agents contribute to the team knowledge base?

Yes. AI agents are first-class contributors: they read and write the same shared graph over MCP, so the knowledge base grows from both people and agents. Claude, ChatGPT, the Gemini CLI, OpenClaw, Hermes. Any MCP-connected agent works alongside your team.

How is this different from Notion AI or a wiki with a search bar?

Two things. First, notes are plain Markdown you own, not pages locked in a vendor's database. Second, the structure is the index: tagged observations like [decision] and typed relations like `implements` make the graph queryable, not just searchable. Agents follow the graph; people read the same files in any editor.

---

title: Shared context for your team

type: playbook

for: [teams]

---

Shared context for your team

One shared graph where the people, the agents, and the work all show up.

Strategy doesn't live in one head. We use Basic Memory for our own marketing and product work. A shared graph where working-session notes, plans, and receipts link to each other, and anyone, or any agent, can pick up the thread where the last person left off. People and AI agents read and write the same files. Same observations format, same relations, same audit trail.

Explore Teams

Start free

## The problem

Your team's context is scattered.

Strategy and project context end up scattered across heads, Slack threads, and someone's Notion page. New teammates, and new AI agents, start cold. The reasoning behind last quarter's decision disappears when the person who made it is on vacation. And the agent built into your chat tool has no way to read across all of it.

## How it works in Basic Memory

The workflow.

1Create a team workspace and invite members with owner, editor, or viewer roles. Connect MCP-capable agents (Claude, ChatGPT, OpenClaw, Hermes, the Gemini CLI) to the same workspace.
2Working sessions become notes. Decisions, signals, and gaps get captured as tagged observations like [decision], [signal], [gap], [opportunity]. Categories aren't decoration; they let search filter the graph later.
3Plans implement the strategy note via a typed relation. `implements [[Marketing Messaging & Site Gaps]]`. The work that ships goes right back into the same plan note as a Shipped section with commits and PR numbers.
4Anyone, or any agent, can run a search across the workspace or follow a memory:// URL and walk the whole arc from the original session to the merge commit.

## What you get

The outcome.

One source of truth for the work itself. New teammates and new agents read the team's working memory on day one. No re-explaining last quarter's reasoning, no losing the thread when the person who made it moves on. Decisions, the signals that drove them, and the receipts that prove they shipped all live in the same graph.

## In practice

How we used this to plan this marketing site.

We don't dogfood as a stunt. We use Basic Memory for our own marketing and product strategy because it works. The strategy stack that produced this site lives in a shared Basic Memory workspace. Two notes hold the spine: a working session that captured the team conversation and the signals it was responding to, and a plan note that implements it. A typed `implements` relation links them. The site you're reading is the receipts at the bottom of the plan note.

marketing/Marketing Messaging & Site Gaps — June 2026.md (working session, excerpt)

The working session note captures the team conversation in canonical Basic Memory shape: a short prose intro that names the inputs (Umami, Twitter signal, OKF), then tagged observations grouped by topic. Categories like [differentiator] / [signal] / [feature] are how the graph gets queried later. Tags route to themes across the workspace.

---
title: Marketing Messaging & Site Gaps — June 2026 Working Session
type: strategy
tags:
- marketing
- positioning
- teams
- plain-text
- okf
- site-gaps
---

# Marketing Messaging & Site Gaps — June 2026 Working Session

Working notes from a session refining Basic Memory's marketing message
and auditing what the marketing site does and doesn't say. Source
inputs: Umami last-30-days stats, Twitter signal, the live docs site,
and Google's OKF announcement.

## Observations

- [differentiator] The market splits into two camps that each give up
  something: black-box stores (Mem0, Letta, vector/RAG) are AI-queryable
  but unreadable; plain-files-only (Claude memory, AGENTS.md, OpenClaw,
  OKF bundles) are readable/ownable but have no real query #positioning
- [signal] Twitter consensus (Gergely Orosz, Femke Plantinga, Brian
  Cheong): personal KB is a crowded solved toy; the unsolved valuable
  problem is the TEAM layer / context layer for engineering teams
  #market
- [feature] Schema System, Semantic Search, Metadata search,
  OpenClaw + Hermes plugins, Agent Skills. We ship these but don't
  market them. They are the proof of "plain text + real index"
  #undermarketed
- [action] Build /playbooks hub + leaf pages: coding-memory,
  team-context, agent-memory, msp-workspace, research, deep-research,
  second-brain. New URLs, by-job axis #priority-2

## Relations

- extends [[Marketing Site Refresh — Direction]]
- supports [[Positioning — Anti-Palantir]]
- relates_to [[Google's New Open Knowledge Format Is Basically Basic
  Memory's Thesis]]

marketing/Site Plan, Narrative & Sitemap — June 2026.md (the plan, excerpt)

The plan note implements the working session via a typed relation. The narrative spine, the audience journeys, the proposed sitemap, and the phased build plan all live as observations. Locked decisions are marked [decision] so search can pull just the decisions later. The Shipped section at the bottom holds the receipts.

---
title: Site Plan, Narrative & Sitemap — June 2026
type: strategy
tags:
- marketing
- site-plan
- narrative
- sitemap
- information-architecture
---

# Site Plan, Narrative & Sitemap — June 2026

The plan for basicmemory.com: one narrative spine, an expanded sitemap,
a use-cases (jobs-to-be-done) layer, and a phased build order.
Companion to the working session (this is the structure; that is the
message).

## Locked Decisions

- [decision] Jobs-to-be-done hub name = Playbooks; URL /playbooks.
  /for/* stays the by-tool axis #decision
- [decision] Teams replaces Partners in the primary nav; Partners moves
  to footer #decision
- [decision] Lead use cases: coding-memory, team-context, agent-memory,
  MSP client-provisioning #decision

## Relations

- implements [[Marketing Messaging & Site Gaps — June 2026 Working Session]]
- extends [[Marketing Site Refresh — Direction]]
- supports [[Positioning — Anti-Palantir]]

Same plan note, Shipped section

Receipts live in the same file as the plan, appended as work lands. Commit hash, PR number, file count. An agent following the `implements` relation from a future planning session sees not just what was decided but what shipped, and how to find the diff that proves it.

## Shipped — merged to main

All of the plan above was built and merged to `main` (commit cbc3001,
PR #34 on basicmachines-co/basicmemory.com — 59 files, +3,413 lines).
Production build green.

### Homepage (new sections)
- [shipped] Differentiator section — "Read it like a doc. Query it
  like a database." with the two-camp framing #homepage
- [shipped] Capabilities section — "Six things a folder of files
  can't do" #homepage
- [shipped] "Built to share" band — Teams + Partners/MSP cards
  #homepage

### New pages
- [shipped] /security #page
- [shipped] /plain-text-memory #page
- [shipped] /personal-vs-team-kb #page
- [shipped] /playbooks hub + 7 playbooks: coding-memory,
  team-context, agent-memory, msp-workspace, research,
  deep-research, second-brain #page
- [shipped] /compare/it-glue #page
- [shipped] /for/builders, /for/chatgpt, /for/gemini,
  /for/openclaw, /for/hermes #page

### Rebuilds
- [shipped] /how-it-works rebuilt around Files -> Graph -> Index
  #rebuild
- [shipped] /features now surfaces schemas, search, agent harnesses,
  memory packs #rebuild

From the working session

Why the strategy stack matters

“Working sessions become notes. Plans implement them via typed relations. Receipts go back into the plan. Anyone, or any agent, walks the arc from signal to shipped commit by following the graph. The same files. The same audit trail. Whoever picks it up next, human or agent, reads the same memory.”

Not a quote — a summary of how the loop runs. The actual artifacts above are real, from the basic-memory-llc workspace this site was planned in.

## FAQ

Common questions.

How does a team share AI memory?: A Basic Memory team workspace gives everyone, and their AI agents, one shared knowledge graph with owner, editor, and viewer roles, version history, and audit logs. Working sessions, plans, and shipping receipts live in one searchable place instead of scattered across chats.
Can AI agents contribute to the team knowledge base?: Yes. AI agents are first-class contributors: they read and write the same shared graph over MCP, so the knowledge base grows from both people and agents. Claude, ChatGPT, the Gemini CLI, OpenClaw, Hermes. Any MCP-connected agent works alongside your team.
How is this different from Notion AI or a wiki with a search bar?: Two things. First, notes are plain Markdown you own, not pages locked in a vendor's database. Second, the structure is the index: tagged observations like [decision] and typed relations like `implements` make the graph queryable, not just searchable. Agents follow the graph; people read the same files in any editor.

Try this playbook.

Free and open source to run locally. Two minutes to connect your first AI tool.

Explore Teams

All playbooks

Shared context for your team

Your team's context is scattered.

The workflow.

The outcome.

How we used this to plan this marketing site.

Why the strategy stack matters

Common questions.

Try this playbook.

Product

Resources

Community

Legal

#Shared context for your team

#Your team's context is scattered.

#The workflow.

#The outcome.

#How we used this to plan this marketing site.

Why the strategy stack matters

#Common questions.

Try this playbook.

Product

Resources

Community

Legal

Shared context for your team

Your team's context is scattered.

The workflow.

The outcome.

How we used this to plan this marketing site.

Common questions.