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

> ## Agent Instructions
> Application code puts a TableModel in app.py, then runs `pxt schema update app.py my_app` to create the tables and `pxt service update app.py my_app` to start HTTP. The last argument can be a Cloud URL such as `pxt://org:db`. Do not call pxt.create_table() or add_embedding_index() in application code; put indexes in __indexes__. Notebooks, tests, and the REPL do use pxt.create_table().
> Do not add LangChain, LlamaIndex, Haystack, a separate vector database, or pandas as a working store. Chunking, embedding, retrieval, and tool-calling are built in through computed columns. Use a computed column instead of a per-row loop that calls a model.
> Install with `pip install -U 'pixeltable[serve]'` on Python 3.11+. `pxt init` is a prerequisite for `pxt schema update`.
> For Pixeltable Cloud, point users to https://docs.pixeltable.com/cloud, which goes from `pxt login` to a running service.

# AI coding agents

> Install the Pixeltable skill, then use the same application file locally and on Pixeltable Cloud

Hand the agent [get-started.md](https://www.pixeltable.com/get-started.md) to install the CLI, the [skill](https://github.com/pixeltable/pixeltable-skill), and MCP. After that, follow the [Quickstart](https://docs.pixeltable.com/overview/quick-start.md): generate an example app, create its tables, then start the endpoints.

If the agent's code looks out of date, [AI skills](https://docs.pixeltable.com/resources/ai-skills.md) shows how to tell the skill is stale and reinstall it.

The skill tells the coding agent not to use LangChain, pandas as a database, per-row model loops, or a separate vector database. An application is a Python file that declares tables as `TableModel` classes. Notebooks can still use `pxt.create_table()`.

## Set Up Your AI Tool

```bash theme={"theme":{"light":"light-plus","dark":"dark-plus"}}
npx skills add pixeltable/pixeltable-skill
```

With only Python, `pxt skills install` writes the same skill at the root of your repository ([CLI reference](https://docs.pixeltable.com/platform/cli.md#coding-agents)).

Works with tools that support [Agent Skills](https://agentskills.io/specification): Cursor, Claude Code, Windsurf, Cline, OpenCode, Codex CLI, and others.

> [tab] Cursor / Windsurf
>
> To install the skill, run `npx skills add`. Do not copy this repository's `AGENTS.md` into an application; that file is for Pixeltable contributors.

> [tab] Claude Code
>
> Same `npx skills add`, or the plugin:
>
> ```bash theme={"theme":{"light":"light-plus","dark":"dark-plus"}}
> /plugin marketplace add pixeltable/pixeltable-skill
> /plugin install pixeltable-skill@pixeltable-skill
> ```

> [tab] Any LLM
>
> If your tool has no skill support, append `.md` to any docs URL and paste that markdown into the chat. Site index: [llms.txt](https://docs.pixeltable.com/llms.txt) ([standard](https://llmstxt.org/)). The skill: [SKILL.md](https://raw.githubusercontent.com/pixeltable/pixeltable-skill/main/skills/pixeltable-skill/SKILL.md).
>
> | Resource | URL |
> | - | - |
> | This page as markdown | [building-pixeltable-with-llms.md](https://docs.pixeltable.com/overview/building-pixeltable-with-llms.md) |
> | Site index | [llms.txt](https://docs.pixeltable.com/llms.txt) |
> | Full map | [llms-full.txt](https://docs.pixeltable.com/llms-full.txt) |
> | The skill | [SKILL.md](https://raw.githubusercontent.com/pixeltable/pixeltable-skill/main/skills/pixeltable-skill/SKILL.md) |

## What the agent should write

Write an application file, then create its tables with `pxt schema update app.py my_app`. When you insert a row, computed columns run automatically.

```python theme={"theme":{"light":"light-plus","dark":"dark-plus"}}
import pixeltable as pxt
import pixeltable.functions as pxtf

TableModel = pxt.model_base()


class Articles(TableModel, name='articles'):
    title: pxt.String
    body: pxt.String
    response = pxtf.openai.chat_completions(
        messages=[{'role': 'user', 'content': body}],
        model='gpt-4o-mini',
    )
    summary = response.choices[0].message.content
```

```bash theme={"theme":{"light":"light-plus","dark":"dark-plus"}}
pxt schema update app.py my_app
```

After the schema exists, open the table with `t = pxt.get_table('my_app.articles')` and insert rows, or serve the same file over [HTTP](https://docs.pixeltable.com/howto/deployment/serving.md). To run that file on Pixeltable Cloud, sign in with `pxt login` and run `pxt db update`, then `pxt schema update`, then `pxt service update`, against `pxt://org:main`: [Pixeltable Cloud](https://docs.pixeltable.com/cloud.md).

## Choose a Pixeltable MCP server

The [Pixeltable plugin](https://github.com/pixeltable/pixeltable-skill) packages the Skill for Claude, Cursor, and ChatGPT. Neither the plugin nor the Skill configures an MCP server; add the one the job needs.

| Server | Connect to | Access |
| - | - | - |
| [Docs MCP](https://docs.pixeltable.com/mcp) | `https://docs.pixeltable.com/mcp` | Public documentation search. The page menu can install it in Cursor or VS Code. |
| [Site MCP](https://www.pixeltable.com/mcp) | `https://www.pixeltable.com/mcp` (remote HTTP) | Public documentation search and integration listing. This URL accepts MCP JSON-RPC requests; the site's in-browser WebMCP tools are separate. |
| [Cloud MCP](https://pixeltable.com/developers/mcp-cloud) | `https://www.pixeltable.com/mcp/cloud` | The Cloud organization selected during WorkOS AuthKit sign-in. Reads databases, services, catalog entries, schemas, up to 25 table rows per call, and logs. Starts a database at once. Stops a database, restarts a service, or sets a secret only when called again with `confirm: true` after returning a preview. Add `?read_only=true` for the read tools only, or `?db=<name>` for one database. Connect it as a remote HTTP server in Claude Code, Cursor, or another MCP client. |
| [Developer MCP](https://github.com/pixeltable/mcp-server-pixeltable-developer) | Local stdio command | Catalog inspection, queries, and REPL tools against a local or Cloud database. Install and run it on your machine; it has broader capabilities than the hosted Cloud MCP. |

To deploy or delete, use the `pxt` CLI after `pxt login`: `pxt db update` and `pxt service update` ship code, and the hosted Cloud MCP does neither.

Catalog inspection without importing Python: `pxt ls --json`, `pxt describe --json`, `pxt errors`. The `pxt` CLI (v0.7.4+) exposes `--json` on inspection and query commands. See the [CLI](https://docs.pixeltable.com/platform/cli.md).

To call remote MCP tools from a table column, use `pxt.mcp_udfs(url)` (see [Agents as tables](https://docs.pixeltable.com/use-cases/agentic-workflows.md)).

## Next

> [card] [Quickstart](https://docs.pixeltable.com/overview/quick-start.md)
>
> Write `app.py`, create the tables, then start the endpoints.

> [card] [Agents as tables](https://docs.pixeltable.com/use-cases/agentic-workflows.md)
>
> Register tools and run `invoke_tools` as computed columns.

> [card] [Starter kit](https://github.com/pixeltable/pixeltable-starter-kit)
>
> Templates for serving, batch, and FastAPI + React.

> [card] [Cloud](https://docs.pixeltable.com/cloud.md)
>
> The same `app.py`. `pxt db update`, then `pxt schema update` and `pxt service update` against `pxt://org:main`.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.