> ## Documentation Index
> Fetch the complete documentation index at: https://docs.correl8.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent system prompt

> Short system prompt so your agent reads Correl8 guidelines once per conversation

Add this block to your agent's system prompt. It tells the model to load project-specific posting rules from Correl8 at the start of each conversation — you do not need to copy the full observations rubric into your prompt.

## Prompt to copy

Replace `{user_id}` with your stable end-user identifier (same value you pass to `post_observations`).

```markdown theme={null}
## Correl8 MCP

At the start of each conversation, call `get_correl8_guidelines(user_id="{user_id}")` — **once only**; **do not call again** if already loaded. It includes observations posting rules and other project-specific Correl8 guidance.
```

## Guidelines tool

The MCP tool **`get_correl8_guidelines`** (pass `user_id`) returns JSON. The **`observations`** and **`research`** fields are markdown describing when and how to call posting tools. Customize them in the app — see [Guidelines](/essentials/using/guidelines).

Same payload via REST (custom integrations): `GET /v1/project/{project_id}/correl8-guidelines/?user_id=…` with `Authorization: ApiKey …`.

## Wiring `user_id` and `interaction_id`

Many teams also inject stable ids into the system prompt or tool context so the model passes them correctly:

* **`user_id`** — your auth or session ID for the end user
* **`interaction_id`** — the current chat thread (e.g. `conv_9f3a2b1c`)

Both are required on every **`post_observations`** call. The model sets **`user_sentiment`** per the loaded guidelines.