Memory for your team and its agents

Work it out once. Every agent finds it next time.

The decisions you take, the lessons you learn and the rules your team agrees are stored once, with who wrote them and who may read them.

Start freeRead the docs
Claude Code · the Mana plugin
  1. /plugin marketplace add mathieubrglt/mana-plugins
  2. /plugin install mana@mana-plugins

Then run /mcp in Claude Code and sign in. There is nothing to paste.

Other agentsCodex, OpenCode, claude.ai, ChatGPT: the steps are in the docs
The Mana mark, a ribbon folded into an M

What one person learns, every agent on the team finds.

An agent stores the lesson during the work. Mana keeps who said it, with which agent, and the day. Mana keeps one memory for your team across every assistant you connect, and the team keeps it when it changes its assistant.

How one lesson moves through the teamTom, in sales, tells ChatGPT why a clinic group was lost, and his agent stores the lesson in Mana. Mana keeps it with its writer and its readers, the whole organization. The next morning, Lena’s claude.ai prepares an onboarding call and Kofi’s Codex writes the site FAQ, and each finds the lesson when it searches. On the topic map, Sales and deals lights its two largest writers, Ines and Tom.TTomSALES · CHATGPTCHATGPT · 22 SEPWe lost the physio group.Procurement asked for our data-processing agreement on day 12.Remember why, for the team.Stored in Mana aslearning/procurement-dpa-before-trialSHARED WITH THE ORGANIZATIONManaPELLWICKlearning/decision-group-pricingconfig/sales/discountslearning/lost-deal-dental-groupTOM · CHATGPT · 22 SEPREADERS · THE ORGANIZATIONLLena · claude.ai23 SEPPrepare the onboarding callfor the 3-site vet group.KKofi · Codex23 SEPWrite the site FAQ from thequestions sales hears.Topic mapPELLWICKSales and dealsPricing and plansCustomer successMarketing and positioningProduct directionHow we workInesTomSALES AND DEALS · INES, TOMlearning/procurement-dpa-before-trialSend the data-processingagreement with the trial,not after itFOUND IN MANA · TOM, 22 SEPSend the data-processingagreement with the trialFOUND IN MANA · TOM, 22 SEPSend the data-processingagreement with the trialHow one lesson moves through the teamTom, in sales, tells ChatGPT why a clinic group was lost, and his agent stores the lesson in Mana. Mana keeps it with its writer and its readers, the whole organization. The next morning, Lena’s claude.ai prepares an onboarding call and Kofi’s Codex writes the site FAQ, and each finds the lesson when it searches. On the topic map, Sales and deals lights its two largest writers, Ines and Tom.TTomSALES · CHATGPTCHATGPT · 22 SEPWe lost the physio group. Procurement askedfor our data-processing agreement on day 12.Remember why, for the team.learning/procurement-dpa-before-trialSend the data-processing agreement with the trialStored in Mana aslearning/procurement-dpa-before-trialManalearning/procurement-dpa-before-trialSend the data-processing agreementwith the trial, not after itTOM · CHATGPT · 22 SEPREADERS · THE ORGANIZATIONLLena · claude.ai23 SEPPrepare the onboardingcall for the 3-site vetgroup.FOUND IN MANA ·TOM, 22 SEPSend the data-processing agreementwith the trialKKofi · Codex23 SEPWrite the site FAQ fromthe questions saleshears.FOUND IN MANA ·TOM, 22 SEPSend the data-processing agreementwith the trialTopic mapSales and dealsPricing and plansCustomer successMarketing and positioningProduct directionHow we workInesTomSALES AND DEALS · INES, TOM

  1. Tom’s agent stores the lesson.

    The physio group asked for the data-processing agreement on day 12 of its trial.

  2. Mana keeps its writer and its readers.

    Tom, with ChatGPT, on 22 Sep. Shared with the organization.

  3. The next morning, two agents find it.

    Lena’s agent prepares a group onboarding call. Kofi’s Codex writes the site FAQ.

  4. The topic map shows who knows.

    Ines and Tom write most of Sales and deals. Ines keeps the procurement pack.

In the demo: Pellwick, an invented company of five people that sells a staff rota app to clinics. Its records were written for the demo.

What changes when your team’s knowledge is kept.

The screens are the real app. Its records and their summaries were written for the demo.

Part of the team map of Pellwick: Tom and Kofi, joined by a dotted line where they share records through a team.
The team map

Know who worked on it before you ask.

The team map shows who wrote what, by person and by agent, over what you may read.

Take the tour

How Mana completes your knowledge base.

A knowledge base holds what people wrote for people. Mana holds what the work taught your team and its agents, with who said it and who may read it.

Your knowledge base and Mana, side by side
QuestionYour knowledge baseConfluence, Notion, SharePointMana
Who writes itPeople, and agents through a connector when someone asks.People and their agents, during the work.
When an agent sees itWhen it searches, through a connector.The rules at session start in Claude Code, with the plugin. In another client, after you paste the prompt of its guide. The rest when it searches.
Who may read itSet for each page, by person or group.Set for each record when it is written: its author, a team, or the organization.

The knowledge base column follows each vendor’s public documentation, read on 1 Oct 2026.

Three steps to a connected agent.

  1. Sign in.

    Create your organization. The first person to sign in is its admin.

  2. Connect an agent.

    In Claude Code, the Mana plugin: two commands and one sign-in.

    Every other agent has its guide in the docs.

  3. Ask it to remember one decision.

    It appears in your Library with your name and the day you said it, and the next session can find it.

Plans

Start free. Pro when your team joins.

One unit is counted: a search, retrieve or answer that runs a model. Reading and storing count nothing.

Free

the default, no subscription, up to 5 members

One organization, every agent. 2,000 model calls a month, 1,000 memories kept. 1 new memory a second is found by meaning at once. Up to 5 topics on the topic map.

Start free

Pro

$20 a month for each member, up to 20 members

50,000 model calls a month, 200,000 memories kept. Realtime indexation: 100 new memories a second are found by meaning at once. Up to 20 topics on the topic map. A 15-day free trial.

Start the trial

How a team works with Mana.

A short series in three parts: how to use it, how a team shares it, and where the time comes back.

All articles
Part 1 of 3 · 9 min read

How to work with Mana

Connect each agent once, and its sessions start with the rules your team keeps. Follow one sales lesson from ChatGPT to claude.ai and Codex.Read the articlePart 2 of 3 · 9 min read

One memory for a team, with clear readers

When people and their agents share one memory, each record has to say who may read it and who may change it. A new account executive's first week at an invented company shows how Mana settles both.Read the articlePart 3 of 3 · 9 min read

Where the time comes back

A team memory gives time back in four places: the context nobody repeats, the approvals nobody taps, the notes that never pass through an agent's model, and the right record reaching the answer. What we measured, and what is still to measure.Read the article

The exact steps for your agent are in the docs. Each guide is also plain Markdown for your agent: add .md to its address.

Take the tourRead the docs