AI Context and Memory

Two related but distinct things shape how the AI behaves across Bitment: the AI Context you write once and that gets read on every AI call, and memory, which is what the AI learns on its own from your conversations and never has to be told twice. Every operator and every firm uses Context; memory works quietly in the background wherever it's available.

Where to find it

The standing AI Context editor lives at the foot of Settings › AI, and the same editor is also reachable from your account dashboard (behind your avatar, under Context). Both show the same content and save the same edits. Agent memory lives on each agent's page, behind the ⋯ menu's Memory item. Chat Project memory lives on a project's page, in the memory rail.

Key concepts

AI Context: a standing prompt, split into a Personal tab (only you see it) and one tab per firm you operate in (shared, edited live by your whole team). It's handed to every AI feature as background.
Memory: short facts the AI distills automatically from your conversations and reuses on future ones. You never write memory directly; it's learned.
Chat Projects: a Claude.ai-Projects-style workspace inside AI Chat that bundles standing instructions, a knowledge base of documents, and a set of conversations. Projects carry their own memory, separate from agent memory.
Agent memory: what one specific agent has learned across its runs and conversations.

Edit the standing AI Context

Open Settings › AI and scroll to the bottom, or go to your account dashboard's Context page.
Use the Personal tab for things only you want the AI to know: "always call me by my first name," "I file on a calendar year." Use a firm's tab for a house style or a shared playbook the whole team's AI should follow.
Edits save as you type. A firm's tab is collaborative: your teammates can be editing the same document at the same time, live.

Whatever you write here is handed to every AI feature as background: AI Chat, the Assistant, and every agent's runs. There's nothing further to configure; it applies automatically the moment you save it.

How it's injected

Every AI call in Bitment reads two standing things before it reads anything else: who it's acting for, and your standing Context (Personal, plus the firm Context of whichever firm the call is billed to). An agent's turn reads the agent's own owning firm's Context; an AI Chat conversation reads whichever firm the conversation belongs to. This is why the firm you pick for a chat or an agent matters beyond billing: it also decides which firm's shared Context applies.

Chat Projects memory

Inside a Chat Project, the assistant also remembers things it learns from your conversations there over time: short facts and preferences it picks up as you chat, not something you teach it directly. This happens automatically in the background after a conversation settles.

Memory is scoped to you personally, even inside a project shared with your whole firm: a teammate's chats build their own memory, not yours.
Memory is on by default. The project's owner can turn it off from the project page with a switch next to the memory list.
You can review everything it's remembered, edit an entry, or delete one you don't want kept, from the memory panel on the project's page.

See AI Chat for the rest of what a Chat Project is.

Agent memory

After an agent's runs and conversations, it distills short durable facts it learned: your preferences, corrections you made, recurring clients and how to handle them, and it reads them on every future wake-up.

Open the agent's page and choose Memory from the ⋯ menu.
Review every fact it has learned.
Rewrite anything it got wrong, or delete anything you don't want kept.
Turn learning off entirely with the Keep learning switch, if you'd rather the agent not retain anything between runs.

Nothing an agent remembers is invisible: this panel is the whole record. Memory is learned automatically; the soul and skills you write are what shape the agent's behavior directly. See AI Agents for the full picture of what an agent is.

What the AI remembers between conversations, and what it doesn't

A brand-new AI Chat conversation starts fresh: it has no memory of your other chats unless it's inside a Chat Project (which carries its own memory) or the conversation @-mentions a client (which pulls in that client's context document as background). An agent's memory is scoped to that one agent; it does not share what it has learned with other agents or with AI Chat. Your standing Context, on the other hand, applies everywhere, every time, because it isn't learned: it's what you deliberately wrote down.

If the AI keeps getting something wrong that isn't a one-off, the fix is usually to write it into your standing Context (durable, deliberate, applies everywhere) rather than hope it eventually learns it (memory is scoped and can take time to form).

Tips

Use Context for anything you'd want every AI feature to already know without being told. Use memory review to catch and correct anything an agent has concluded that isn't quite right.
A firm's Context tab is a good place to put your house style for client communication: it applies to every agent and every chat billed to that firm.
If an agent seems to have the wrong idea about a recurring situation, check its Memory panel before rewriting its soul; the fix might just be one wrong fact.