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

# Build with a coding agent

> Point Claude Code, Cursor, Codex or any coding agent at one file and it can build against the Skylit API and MCP server.

Skylit publishes a single, agent-ready guide: every endpoint, parameter, limit
and error rule in one Markdown file, generated from the same OpenAPI specs as
this reference. Give it to your coding agent and ask for what you want.

|                  |                                                                                                                                                                                                         |
| ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Agent guide**  | `https://docs.skylit.ai/.well-known/agent-skills/skylit-api/skill.md`                                                                                                                                   |
| **Skills index** | `https://docs.skylit.ai/.well-known/agent-skills/index.json`                                                                                                                                            |
| **Docs index**   | `https://docs.skylit.ai/llms.txt` (append `.md` to any page URL for Markdown)                                                                                                                           |
| **OpenAPI**      | [`openapi.yaml`](https://docs.skylit.ai/openapi.yaml) · [`flowseeker-openapi.yaml`](https://docs.skylit.ai/flowseeker-openapi.yaml) · [`atlas-openapi.yaml`](https://docs.skylit.ai/atlas-openapi.yaml) |

## Install it as a skill

Agents that support skills can install the guide straight from the docs site:

```bash theme={null}
npx skills add https://docs.skylit.ai
```

## Or paste this prompt

Works with any agent that can read a URL. Set `SKYLIT_API_KEY` in your
environment first (create a key on the
[Developer page](https://app.skylit.ai/developer)).

```text theme={null}
Read https://docs.skylit.ai/.well-known/agent-skills/skylit-api/skill.md and follow it exactly.
Use my key from the SKYLIT_API_KEY environment variable with the header
"Authorization: Bearer <key>". Then build: <describe what you want, e.g.
"a Python script that streams SPY dealer positioning and prints the king node
every time it changes">.
```

## Or connect the MCP server

If your agent should query Skylit while it works rather than write code
against it, connect the MCP server instead:

```bash theme={null}
claude mcp add --transport http skylit https://mcp.skylit.ai/mcp \
  --header "Authorization: Bearer $SKYLIT_API_KEY"
```

Other clients are covered in the [MCP quickstart](/mcp/quickstart).

<Note>
  The guide covers the rules agents most often get wrong: bearer auth only, up
  to 10 symbols per heatmap call, one symbol per stream, the `layout=matrix`
  grid, the historical concurrency limit, and which errors to retry.
</Note>

<Warning>
  API access is licensed to you for your own trading and research. An agent
  acting on your key is bound by the same terms: no redistribution, public
  bots or model training on Skylit data.
</Warning>