A coding agent that can read your database, errors and analytics
Connect a service your workspace already uses, such as Neon, Supabase, PostHog or Sentry, and every Sinatra agent in the workspace can use it. The agent can check a schema, look up an error, or read the numbers behind a bug report before it writes the fix, instead of guessing.
What a connection gives an agent
Each connection reaches the provider's own hosted MCP server. MCP (Model Context Protocol) is a standard way for an AI agent to call tools that a service publishes, so the tools an agent gets are the ones the provider built. Once a service is connected, every agent run in the workspace gets those tools, on tasks from Linear, GitHub, Slack, the CLI, the API or MCP alike. Agents can use a new connection from their next run.
When a task would have gone better with a service you haven't connected, the agent can end its reply with a link to connect it. It finishes what it can first, says what it couldn't check, and offers each service at most once in a conversation. Automated fixes, such as reworking a failing CI check, don't offer links. Connecting from a task started in the CLI, MCP or API picks that task back up; for a task started anywhere else, reply on the task to continue.
The services you can connect
Eight services, each through its own MCP server. What actually limits an agent depends on the service and on how you connect it.
| Service | How you connect | What limits the agent |
|---|---|---|
| Neon | Sign in, or paste an API key | Sign-in: read-only on one project, enforced by Neon. Token: read-only requested, not guaranteed. |
| Supabase | Sign in, or paste an access token | Read-only requested, not guaranteed, either way. |
| PostHog | Sign in, or paste a personal API key | Sign-in: read-only, enforced by PostHog. Token: the key's own permissions. |
| Sentry | Sign in, or paste a personal token | The skills you leave ticked, or the token's scopes. |
| PlanetScale | Sign in, or paste a service token | The tools and databases you tick, or the token's permissions. |
| Cloudflare | Sign in, or paste an API token | The access you approve, or the token's permissions. |
| Notion | Sign in, or paste an MCP token | Your full Notion permissions. Notion has no read-only option. |
| Vercel | Sign in only | Your Vercel access in the teams you share. Vercel has no read-only option. |
Signing in sends you to the provider to approve access, and the provider sends you back. Pasting a token takes one you create at the provider; the connect dialog links to the page where it's created and says which permissions to give it. What to tick or untick at each provider is in the connectors docs.
What an agent can do with a connection
We don't make these connections read-only. Sinatra tells agents to use connected services for reading only and to treat what comes back as untrusted data, but that is an instruction to the model, and it doesn't stop anything. What an agent can actually do is set by the token you paste or what you approve at the provider. Every connection carries one of three labels:
Read-only, enforced by the provider. The provider refuses changes for this connection. Only Neon's and PostHog's sign-ins get this label.
Read-only not guaranteed. Sinatra asks for read-only access with a flag in the MCP address, and an agent can remove that flag. Neon tokens and Supabase connections get this label.
May be able to make changes. Sinatra adds no limit. Whatever the token or sign-in allows, an agent can do.
Before a connection without enforced read-only is saved, a warning lists what an agent could do if the token or sign-in allows it. For Supabase that is running SQL that changes or deletes data, applying migrations, deploying Edge Functions, and reading every row, your logs and your project's keys. Then you tick each confirmation separately. Every connection's text says that agents can use the service on every task in the workspace, for every member; that what they read is sent to the AI model running the task and can include stored tokens and keys; that content the agent reads can influence what it does; and that granting only the access you intend is your responsibility. Unless the provider enforces read-only, you also confirm that an agent may change or delete things. Sinatra records who accepted, when, and which version of the text. Neon recommends connecting agents to development and testing data, not production.
How the token stays out of the sandbox
The agent runs in its own cloud sandbox, and the token never goes in. Sinatra stores it encrypted with a key bound to your workspace, and the agent's MCP configuration holds only a placeholder. When the agent calls the provider, Sinatra's egress proxy swaps the placeholder for the real token only on a TLS request to that provider's MCP host, such as mcp.neon.tech, only in the request's Authorization header, and only while a run that was given the connection holds it. A placeholder sent anywhere else isn't swapped.
A pasted token is sent once and never shown again. Responses from the API, the CLI and MCP never include a token or any part of one. How the rest of the sandbox is isolated is on the security page.
Connecting a service
A workspace owner or admin signs in at app.sinatra.dev and opens Settings, then Connectors. Members can see the connections but can't add or remove them.
Select Connect next to the service, then choose Sign in or Paste a token where the service offers both. Fill in the provider's reference if it takes one, such as Neon's project ID, and optionally a name such as production.
Read Before you connect and tick each confirmation. The text depends on the service, how you connect and the access label.
A sign-in takes you to the provider and back to Settings, which says whether it worked. A token is saved when you select Connect. Agents can use the connection from their next run.
From the CLI, the API or MCP
You can also manage connections with sinatra connectors, the API and Sinatra's MCP server. None of them takes a token as a command-line or tool argument, and none completes a sign-in. For a service you can sign in to, the CLI and MCP give you a link to the dashboard's connect dialog, filled in with the project and name, where an owner or admin signs in and accepts the warning. To paste a token from the CLI, pass --token: it shows the warning and confirmations, then reads the token at a hidden prompt or from stdin. The API connects a pasted token sent in the request body along with the confirmations you accepted. The commands and endpoints are in the connectors docs.
What it costs
Connectors are on for every workspace, Free included. Free is 1 agent at a time, plus 1 waiting, with no daily limit and no card. Pro is $20 a month for the whole workspace: 3 agents at a time, plus 10 waiting, any number of people. Bring your own Claude, ChatGPT or Cursor subscription, or an API key. Sinatra never bills for model usage. That's between you and your provider.
Common questions
Are connections read-only?
Not in general, and Sinatra doesn't enforce it. Only Neon's and PostHog's sign-ins have the provider refusing changes. Neon tokens and Supabase connections ask for read-only access with a flag that an agent can remove. Everywhere else an agent can do whatever the token or sign-in allows. Agents are told to use connected services for reading only, but that is an instruction to the model and doesn't stop anything. Each connection is labelled with which case it is.
Does the token go into the sandbox?
No. Sinatra stores it encrypted with a key bound to your workspace, and the agent's MCP configuration holds only a placeholder. Sinatra's egress proxy puts the real token in the Authorization header of TLS requests to that provider's MCP host, and nowhere else. Responses from the API, the CLI and MCP never include a token or any part of one.
Who can connect a service?
A workspace owner or admin. Every member can see the workspace's connections, and every task a member starts can use them. A connection keeps working only while the person who made it is an owner or admin; if they leave or become a member, Sinatra disconnects the connections they made but doesn't revoke them at the provider.
Can I connect the same service twice?
Yes. A workspace can have up to 20 connections, including more than one to the same service, for example one per Neon project. Give each a name, such as production, so you can tell them apart.
What happens when I disconnect?
Agents lose access on their next call and Sinatra deletes its copy of the token. A sign-in is revoked at the provider. A pasted token keeps working at the provider until you revoke it there, which Sinatra can't do for you. Cloudflare, Sentry and Vercel also keep their MCP server listed as an authorized application after a sign-in is disconnected, so remove it in your account settings there.
Do sign-ins expire?
Some do. Sentry and Cloudflare end a sign-in 30 days after you approve it, and Notion after 180. The connection then shows Needs sign-in again and agents don't get it until an owner or admin signs in again. A pasted token isn't refreshed: if it expires or you revoke it, the agent's calls fail until you connect a new one.
Which plans include connectors?
Connectors are on for every workspace, including Free: 1 agent at a time, plus 1 waiting, no card. Bring your own Claude, ChatGPT or Cursor subscription, or an API key. Sinatra never bills for model usage. That's between you and your provider.
Connected services work on tasks from a Linear assignment, the Sinatra label on a GitHub issue, a Slack mention, an MCP client or the CLI and API. Questions about connectors go to contact@sinatra.dev, or see Support.