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

# sync_icp

> Sync the user's existing ICP and context files into the graph (file to graph). Despite the name it is a sync, not a read.

Syncs the user's existing ICP and positioning files into Nous, instead of making them re-author their ICP in a second place. The name reads like a getter, but honestly it is a **sync**: you read the user's own markdown files with your file tools and push each one into the graph. Nous keeps a served copy of the prose, rebuilds the ICP scoring model from it, and remembers each file's path so [`export_icp_model`](/mcp/tools/export-icp-model) can later write the learned model back into the same file.

<Note>Direction is **file to graph**. The user's files stay the source of truth for the prose. Nous owns the learned scoring half. The reverse direction (graph to file) is [`export_icp_model`](/mcp/tools/export-icp-model).</Note>

## Parameters

| Name                     | Type      | Required | Description                                                                                                                                         |
| ------------------------ | --------- | -------- | --------------------------------------------------------------------------------------------------------------------------------------------------- |
| `sections`               | object\[] | ✓        | One entry per ICP or positioning file (or section) you read.                                                                                        |
| `sections[].section`     | enum      | ✓        | Which section this maps to: `ICP`, `Market`, `Product`, `Pricing`, `Competitors`, `Positioning`, `GTM Motion`, or `Notes`.                          |
| `sections[].content`     | string    | ✓        | The section's content, read from the file (trimmed prose, not the whole repo).                                                                      |
| `sections[].source_path` | string    |          | The file this came from, relative to the project root, e.g. `context/icp.md`. Required on the ICP section so the learned model can be written back. |

Map granularly: one file to the single section it best fits (`icp.md` to ICP, `positioning.md` to Positioning, `pricing.md` to Pricing), one entry per file. If one file holds several sections under headers, split it by header into multiple entries. Keep any "not a fit" exclusions inside the ICP section, because Nous turns each stated exclusion into a hard disqualifier that caps those accounts below Not-ICP.

## Returns

```
Synced 3 sections from the user's files:
  • ICP          ← context/icp.md
  • Positioning  ← context/positioning.md
  • Pricing      ← context/pricing.md

Built the ICP scoring model (4 signals).
Next: if the user can name a few closed-won + closed-lost domains, call train_icp_model to sharpen it on real outcomes, then call export_icp_model to write the learned model back into their ICP file.
```

If there's nothing valid to sync:

```
Nothing to sync. Read the user's ICP/positioning file(s) first and pass each as a section (ICP, Positioning, Pricing, …) with its source_path.
```

## When to call it

* **During onboarding** → find the project's existing GTM files (`context/`, `gtm/`, files named `icp*`, `positioning*`, `pricing*`, `competitors*`), read them, and sync them
* **Whenever an ICP or context file changes** → re-sync in the same turn, because the edit does not change the score, the exclusions, or the model until you do

To write the learned model back into the file, use [`export_icp_model`](/mcp/tools/export-icp-model). To sharpen the model on real outcomes, use [`train_icp_model`](/mcp/tools/train-icp-model).

## Backed by

`POST /v2/workspace/icp/import`. The agent setup surface. Shares its implementation with the web GTM Context page.