> For the complete documentation index, see [llms.txt](https://docs.countercyclical.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.countercyclical.io/agents/context.md).

# Context

Define persistent background information and standing guidance that Agents should consider when working for you or your workspace.

Context gives Agents information that should remain available across runs without requiring you to repeat it in every prompt. Use it for durable background, preferences, terminology, operating assumptions, and other guidance that should consistently inform Agent work.

See also [Conversations](https://docs.countercyclical.io/agents/conversations) and [Agent Projects](https://docs.countercyclical.io/agents/projects).

Countercyclical supports two context scopes: **Personal Context**, which applies to your Agent runs, and **Workspace Context**, which applies across the workspace.

### Personal Context

Personal Context contains background and standing guidance associated with your account.

To manage it, go to [**Settings › Account › Agents › Context**](https://dashboard.countercyclical.io/settings/account/agents/context).

Personal Context is useful for information such as:

* Your role, responsibilities, and areas of focus.
* Recurring preferences for analysis or communication.
* Standing assumptions Agents should use when working for you.
* Terminology, conventions, or priorities specific to your work.
* General guidance that should accompany your selected skills and references.

Enter your guidance in the **Personal Context** field and select **Save Context**.

Personal Context supports up to **5,000 characters**.

Once saved, it is applied to your Agent runs in addition to the skills, references, and other context selected for a particular request.

#### What to Put in Personal Context

Personal Context should contain information that is broadly useful across many Agent runs rather than instructions for one specific task.

For example, an investment professional might include their investment mandate, preferred analytical conventions, typical holding period, recurring definitions, or expectations for how uncertainty should be communicated.

Avoid putting temporary instructions such as “analyze this company” or “prepare a memo for Friday” into Personal Context. Those instructions belong in the Agent prompt itself.

### Workspace Context

Workspace Context contains shared background and standing guidance for Agents operating in a particular workspace.

To manage it, go to [**Settings › Workspace › Agents › Context**](https://dashboard.countercyclical.io/settings/workspace/agents/context).

Workspace Context is useful for information that should apply consistently across the organization or team, such as shared investment mandates, research standards, internal terminology, portfolio assumptions, organizational conventions, or standing analytical guidance.

Enter the shared guidance in the **Workspace Context** field and select **Save Context**.

Workspace Context supports up to **5,000 characters**.

Once saved, the context is applied to Agent runs across the workspace.

### Personal Context and Workspace Context

Personal and Workspace Context can be used together.

**Workspace Context** establishes the shared background for Agent work in the workspace. **Personal Context** adds information specific to you and your Agent runs.

This allows an organization to maintain common assumptions while individual users add their own working preferences and background.

Workspace governance remains authoritative. Personal Context cannot override workspace governance controls.

### Context vs. Skills

Context and Skills serve different purposes.

Use **Context** for information Agents should generally know. Use **Skills** for reusable instructions describing how Agents should perform particular kinds of work.

For example, an investment mandate such as “we focus on North American vertical-market software businesses” belongs naturally in Context. A detailed procedure for evaluating SaaS retention metrics belongs more naturally in a Skill.

A useful rule is:

**Context describes what Agents should know. Skills describe how Agents should work.**

Keeping this distinction makes both easier to maintain and reduces unnecessary instructions in individual prompts.

### Using Context in Agent Runs

Saved context is applied automatically according to its scope. You do not need to manually attach it each time you start an Agent run.

You can still provide run-specific context directly in the composer. Prompt instructions, attached materials, selected skills, and other references can supply information specific to the current task while your saved Context provides the persistent background.

This makes Context particularly useful for assumptions or preferences that would otherwise need to be repeated across many requests.

### Referencing People with @ Mentions

The Agents composer supports **@ mentions** for adding relevant people & workspace resources as context to a request.

Type `@` in the composer to open the available workspace menu, then select the person or resource you want to reference. The selected item(s) can then be incorporated into the context of that Agent request.

This is useful when the work depends on a particular person or artifact, such as preparing analysis for a stakeholder, incorporating information associated with a collaborator, or making it clear who a request concerns.

Unlike saved Personal or Workspace Context, an `@` mention is **run-specific**. Use it when a person or workspace resource matters to the current request rather than as permanent standing guidance.

### Writing Effective Context

Keep saved Context durable and broadly applicable. Good Context should continue to make sense weeks or months later without depending on a particular conversation.

Prefer concise statements of facts, conventions, preferences, and standing guidance. Separate unrelated concepts where useful so that Agents can interpret them cleanly.

For example:

> We invest primarily in profitable or near-profitable B2B software businesses. When discussing valuation, distinguish reported multiples from forward estimates. Flag assumptions explicitly when primary-source evidence is unavailable. Use USD unless another currency is material to the analysis.

This provides reusable background without prescribing an entire workflow.

If you find yourself describing a multi-step procedure, required output structure, or specialized methodology, consider creating a **Skill** instead.

### Updating Context

Update Personal or Workspace Context when the underlying background or standing guidance changes.

Because Context is designed to persist across Agent runs, remove assumptions that are no longer valid rather than relying on later prompts to continually correct them.

For Workspace Context, changes can affect Agent work across the workspace. Treat material updates as changes to shared operating guidance and review them accordingly.

### Governance and Context

{% hint style="info" %}
Context influences how Agents interpret and perform work, but it does not grant additional capabilities or permissions.
{% endhint %}

Workspace governance continues to control which models, tools, data sources, and governed actions Agents may use. Context cannot be used to bypass those restrictions.

This separation allows organizations to provide rich standing guidance while maintaining centrally enforced controls over Agent execution.

#### Related Pages

{% content-ref url="/pages/ikJo8G678beqJUbsGBMH" %}
[Skills](https://docs.countercyclical.io/agents/skills.md)
{% endcontent-ref %}

{% content-ref url="/pages/Ff3W2jW4IMpZKEEf5yRi" %}
[Conversations](https://docs.countercyclical.io/agents/conversations.md)
{% endcontent-ref %}

{% content-ref url="/pages/Y4cgFnb6MdXOEYWqzLwJ" %}
[Projects](https://docs.countercyclical.io/agents/projects.md)
{% endcontent-ref %}
