# Context

The workspace agent context is a short briefing about your product that the PM agent, the
interview agent and some generators read: what the product is, who uses it, the words your team uses
and the conventions that always apply. Think of it as what you would tell a new colleague on their first
day. You write it once on the **Context** tab, or generate it from your product overview.

<Availability permission="Agents access on the Roles tab, under Administrative access. The Owner, Admin, Power User and Developer roles have it." />

## Key concepts

**Workspace agent context.** One Markdown document per workspace. Productmind adds it to the
instructions of the
[PM agent](/user-guide/get-started/finding-your-way-around#the-pm-agent), and of some of
the features that generate content for you, such as generating tasks from a feature or promoting
feature requests. It also reaches the agent that runs
[agentic interviews](/user-guide/feedback/agentic-interviews) with your customers, and the agents
that summarize your customer research.

**Product overview.** The overview document of your [knowledge base](/user-guide/knowledge-base).
Productmind can write the context for you from it.

**Context or skills.** The agents that read the context read it on every run, so keep it to
what is always true. A
procedure that only matters for one job belongs in a [skill](/user-guide/engineering/agents/skills).

## The Context tab at a glance

The **Context** tab isn't shown in the **Agents** tab list for now. A link to it still opens it,
and the agents that read the context still get the one your workspace saved.

<Screenshot id="agents-context" alt="The Context tab: the Workspace agent context heading, the Generate from product overview button and an editor" />

The tab shows the **Workspace agent context** heading, **Generate from product overview**, and
the context in an editor. Below the editor, **Injected up to ~4KB per agent run** is a reminder of
how much of it agents read.

## Writing the context

### To write the context

1. Open the **Context** tab.
2. Write the context in the editor. For example: what the product does and for whom, the terms
   your team uses and what they mean, and the conventions every answer should follow.
3. Move out of the editor.

The context saves, and the header shows **Saved**. Agents read the new version within about five
minutes.

### To generate the context from your product overview

1. On the **Context** tab, choose **Generate from product overview**.
2. If the editor already holds text, Productmind asks you to confirm. Choose **Replace context**,
   or **Keep current context** to cancel.
3. Wait while Productmind reads your product overview and writes a briefing from it. The button
   shows **Generating** meanwhile.

The generated context appears in the editor. Review it and edit it like any other text.

:::caution
Generating replaces the context in the editor, including your edits, and the old text can't be
recovered. Copy anything you want to keep before you choose **Replace context**.
:::

## How the context behaves

**It's written for you the first time.** When your workspace saves its product overview in the
knowledge base and has no context yet, Productmind generates one from it. It never overwrites a
context that already exists.

**Agents read about 4,000 characters of it.** Anything longer is cut off, so put the most important
things first.

**Changes reach agents within about five minutes.** Agents keep a copy of the context for a few
minutes between runs.

**The interview agent reads it too.** Don't put anything in the context that you wouldn't want
the agent that interviews your customers to know, or to repeat to them.

**Coding sessions, scans and the agents your automations run don't read it.** To shape how
coding sessions and scans work, edit their [templates](/user-guide/engineering/agents/templates)
or add [skills](/user-guide/engineering/agents/skills). An agent that an automation runs follows
only its own instructions, which you edit on the
[Definitions](/user-guide/engineering/agents/definitions) tab.

## Permissions

| To | You need |
| --- | --- |
| Write or generate the context | **Agents** access (under **Administrative access** on the **Roles** tab) |

The **Owner**, **Admin**, **Power User** and **Developer** roles have it. The PM agent, the
interview agent and the generators read the context whoever uses them. See
[Roles and permissions](/user-guide/reference/roles-and-permissions).

## Troubleshooting

### Agents ignore part of the context

Agents read only about the first 4,000 characters. Shorten the context, and move procedures into
skills.

### Generate from product overview produced little or nothing

The context is only as good as your product overview. Fill in the overview in the knowledge base,
then generate again.

## Related topics

- [Agents](/user-guide/engineering/agents/overview)
- [Skills](/user-guide/engineering/agents/skills)
- [Knowledge base](/user-guide/knowledge-base)
