MCP server for Claude
Point Claude at your database. Get a dashboard your team can open.
Claude Code and claude.ai can work inside Fastero on your behalf — run SQL on your live data, save the queries, and leave the result somewhere your colleagues open next week. Not a chat transcript. One install line, and a key that acts as you in one organization.
Install
One line in Claude Code.
Create a key in Fastero and paste the command it gives you. It works the same in bash, zsh, PowerShell and cmd.
claude mcp add --scope user --transport http fastero https://api.fastero.com/api/v1/mcp --header "Authorization: Bearer fst_pat_..."Prefer claude.ai? Add https://api.fastero.com/api/v1/mcp as a custom connector, sign in, and choose your organization.
Create a key
In Fastero: Settings → Developers → AI Agent Access → Create key. Choose "Build dashboards" so the key may connect a database and create dashboards. The key is shown once, with the command below ready to paste.
Add the server
Paste one line into your terminal. /mcp in Claude Code should then show fastero · connected. On claude.ai, add the same URL as a custom connector and sign in instead.
Point it at your data
Ask Claude to connect your database, or to use one you have already connected. It runs SQL through Fastero, so credentials stay in Fastero and never reach the conversation.
Ask for a dashboard
Claude saves the queries and builds the widgets. Open the dashboard in Fastero, invite a colleague, or publish a share link for people with no account.
What Claude can do
Six kinds of work, switched on one at a time.
Each group needs its own permission, ticked when you create the key. A read-only key sees only the tools it can call — it cannot even list the rest.
Explore what you have
List your data connections and projects, read a connection's tables and columns, and run SQL against any of them.
Read · Run SQL
Connect a database
Create a connection from credentials you hand it: Postgres, MySQL, SQL Server, Redshift, Snowflake or Oracle. The connection is tested before it is saved.
Create connections
Build dashboards and reports
Save a query, turn it into a table, a number or a chart, and place widgets on a dashboard your team opens in the browser.
Read · Create and update
Notebooks, Python and files
Create and write Jupyter notebooks, create Python projects, write their files and run them, and read what a run produced.
Create and update · Run
Secrets and schedules
Set the credentials a Python project runs with — values go in and are never read back — and turn scheduled runs on or off.
Secrets · Read and update
Clean up
Delete a connection, query, report, dashboard, notebook, Python project or file. Off unless you switch it on.
Delete
Permissions
The key acts as you. Never more.
Your role decides
Every call runs the same permission checks as the app itself, against your live role. There is no second, weaker path to your data.
One organization per key
A key is bound to the organization you chose when you made it. Working in two organizations means two keys, which can both be connected at once.
Revoke, and it is over
Revoking a key stops it immediately, and revoking a claude.ai connector ends that session rather than letting it renew. Keys last at most 90 days.
Credentials stay put
Connection credentials are stored encrypted in Fastero and are never handed back to the agent. Secret values can be set through the agent, never read.
Every call is logged
Each tool call is written to your organization's audit log with the key that made it. Arguments are deliberately not recorded: they carry SQL and credentials.
You check the numbers
An agent writing SQL can be wrong, exactly as a person can. Read what it built before you send it round: the queries behind every widget are there to open.
Questions people ask first
What can the key do?
Exactly what you tick when you create it, and never more than your own role in that organization allows. Permissions are checked against your live role on every call, so if your access changes later, existing keys narrow automatically. A key belongs to one organization; a second organization means a second key.
Does Claude see my database password?
Only if you paste it into the conversation to create a connection. Once a connection exists in Fastero, Claude uses it by id: it asks Fastero to run SQL and gets rows back, and the stored credentials are never returned to it.
How do I switch it off?
Settings → Developers → AI Agent Access, and revoke the key. It stops working immediately, and revoking a claude.ai connector ends that connection too rather than letting it renew. Keys last at most 90 days, and you get a reminder 14 days and 3 days before one expires.
Can I use it with claude.ai instead of Claude Code?
Yes. Add https://api.fastero.com/api/v1/mcp as a custom connector in claude.ai, sign in, choose the organization, and approve the permissions. Same server, same permissions; no key to paste.
Do my colleagues need Claude to see the result?
No. What Claude builds is a normal Fastero dashboard. Colleagues open it after signing in, and a public share link works for people with no account at all.
Will the dashboard update by itself?
A dashboard you open in Fastero runs its queries against your database. A public share link is a snapshot taken when you shared it, so republish it when you want it to show newer numbers.
Give your agent somewhere to put the work.
Create a key, paste one line, and ask Claude for a dashboard.
Create an agent key