---
name: sequel
description: Use this skill for any question about the user's business data or connected accounts, including databases and warehouses, product analytics, revenue and CRM, and SEO (Search Console, GA4, Ahrefs, rankings, traffic drops, indexing, competitors, backlinks). Sequel connects the account once and exposes it over MCP, and ships playbooks the agent reads with sequel_read_skill so it follows a proven procedure instead of improvising.
version: 1.1.0
homepage: https://sequel.sh
---

# Query your data with Sequel

Sequel is the data layer for AI agents. It manages authentication, credentials,
and query execution so an agent can securely query cloud databases, data
warehouses, product analytics, and SaaS tools over the Model Context Protocol
(MCP). You connect a source once in Sequel; your agent talks to Sequel, and
Sequel talks to your data. Adding a new tool becomes a permission grant rather
than a secret-sharing exercise.

## Connect the MCP server

Sequel exposes a single streamable-HTTP MCP endpoint:

- **Endpoint:** `https://api.sequel.sh/mcp`
- **Auth:** `Authorization: Bearer <api_key>` (keys look like `sql_...`), or OAuth
  (the server advertises RFC 9728 / RFC 8414 metadata under
  `https://api.sequel.sh/.well-known/`).

Example install for Claude Code:

```bash
claude mcp add --transport http sequel https://api.sequel.sh/mcp --header "Authorization: Bearer sql_your_api_key"
```

Setup snippets for Cursor, VS Code, Windsurf, ChatGPT, Codex, Zed, Gemini and
more are at https://sequel.sh/docs/install.

## Tools

| Tool | Purpose |
| --- | --- |
| `sequel_manage_connections` | List your org's connections and start connecting a new source. |
| `sequel_search` | Discover available data sources and fetch an execution plan. Mints a `session_id`. |
| `sequel_execute` | Run a one-off query and write the result to a file. |
| `sequel_workbench` | Python/Bash sandbox (pandas, numpy, matplotlib; `import sequel`) for analysis. |
| `sequel_download_file` | Get a signed URL for a file produced during the session. |
| `sequel_list_skills` | Every playbook as id and name. Use when search did not surface one. |
| `sequel_read_skill` | Read a playbook by id or slug. Returns its steps plus which of its apps are connected. |

## Typical flow

1. Call `sequel_search` to find the relevant source and get a plan. Keep the
   returned `session_id`.
2. Thread that `session_id` into `sequel_execute` and `sequel_workbench` so they
   share the same `/data` working context and file outputs.
3. Use `sequel_download_file` to retrieve any generated artifacts.

## Skills: follow the playbook, do not improvise

`sequel_search` also returns a `skills` array: curated playbooks for the kind of
question being asked. If one matches, call `sequel_read_skill` with its
`skill_id` before designing your own approach, and follow it. A playbook names
the exact tools to call, in order, with the parameters and the gotchas that
otherwise produce a confidently wrong answer.

Playbooks can hand off to each other by slug; `sequel_read_skill` accepts a
slug as well as an id. The result also lists the apps the playbook is written
for and whether each is connected. Apps are advisory: run the steps the
connected sources support, say which were skipped, and offer to connect a
missing one with `sequel_manage_connections` (`action: "connect"`) rather
than asking for credentials in chat.

For anything about SEO or organic search, the entry playbook is the slug
`seo`: read it first. It connects Search Console, routes to the focused
playbook (traffic drops, striking-distance keywords, cannibalization, indexing
and sitemaps, competitors and backlinks, technical audits, AI search
visibility, organic revenue), and states the evidence rules every SEO answer
follows. The full catalogue is browsable at https://sequel.sh/skills.

Sequel supplies account data. Fetching live pages, robots.txt or a sitemap is
done with your own browsing tools when a playbook calls for it.

Sequel connects to cloud-hosted sources (e.g. RDS, Supabase, Neon, BigQuery,
Snowflake, ClickHouse, Google Analytics). Full documentation: https://sequel.sh/docs;
every page is also available as raw Markdown by appending `.md` to its URL.