TinyFish MCP integration: operations for AI agents
TinyFish MCP connects to Type.com as an integration. This page lists the 56 operations documented for Claude and Codex in Type.com: 24 read, 21 write, 11 destructive.
About TinyFish MCP
AI-powered browser automation and web interaction agent. Enables AI to browse, interact with websites, and perform web tasks.
Categories: Developer Tools & IT
How to connect TinyFish MCP to Type.com
- Access
- Organization connection: one shared account for the workspace.
To add the connection, follow the steps in Space connections.
Documented operations
56 operations. Your connection's permissions determine which operations are available. API references describe the vendor surface; they do not expand your connection's credentials or access.
56 operations listed.
Automation
- Write
Run browser automation with SSE streaming
Execute a browser automation task with Server-Sent Events (SSE) streaming. Returns a real-time event stream with automation progress, browser streaming URL, and final results.
- Write
Run browser automation synchronously
Execute a browser automation task synchronously and wait for completion. Returns the final result once the automation finishes (success or failure). Use this endpoint when you need the complete result in a single response. Note: Runs created via this endpoint cannot be cancelled.
- Write
Start automation asynchronously
Creates and enqueues an automation run, returning the runid immediately without waiting for completion. Use this for long-running automations where you want to poll for results separately.
- Write
Start multiple automations asynchronously
Creates and enqueues multiple automation runs in a single request, returning runids immediately without waiting for completion. Maximum 100 runs per request. Atomic creation: Run creation is all-or-nothing. Either all runs are created successfully, or none are (returns error).
Runs
- Read
List and search runs
List automation runs with optional filtering by status, goal text, and date range. Returns paginated results with total count. Default sort order is newest first.
- Read
Get multiple runs by IDs
Retrieve multiple runs by their IDs in a single request. Returns found runs and lists any IDs that were not found or not owned. Maximum 100 IDs per request.
- Read
Get run by ID
Get detailed information about a specific automation run by its ID.
- Read
Get step screenshot
Serve the JPEG screenshot captured at a specific step via API key. Returns binary image data. Note: screenshots=url on GET /v1/runs/{id} returns browser-accessible URLs at /runs/{id}/steps/{stepId}/screenshot (session auth).
- Read
Get step HTML snapshot
Serve the full-page HTML snapshot captured at a specific step via API key. Returns raw HTML. Only available when the run was started with captureconfig.html: true.
- Destructive
Cancel multiple runs by IDs
Cancel multiple runs by their IDs in a single request. Returns per-run results including cancelled runs, already-terminal runs, and not-found IDs. Maximum 100 IDs per request. Idempotent: calling twice returns consistent results.
- Destructive
Cancel run by ID
Cancel a run by ID. Only runs created via /v1/automation/run-async or /v1/automation/run-sse can be cancelled. Runs created via the synchronous /v1/automation/run endpoint cannot be cancelled.
Browser
- Read
List browser session usage
List Tetra browser session usage for the authenticated user. Returns session telemetry including duration, data transfer, mode, and status.
- Write
Create a remote browser session
Creates a remote tf-browser session and returns CDP connection details. Connect to the browser via CDP WebSocket at cdpurl. By default the session egresses through the TinyFish residential proxy in the US, pinned to one exit IP for the life of the session.
- Destructive
Terminate a browser session
Terminates a remote browser session immediately. If the session is already ended this call is idempotent and still returns 204.
Fetch
- Read
Fetch and extract content from URLs
Fetches web pages, renders JavaScript-heavy pages when needed, and returns clean extracted content in your preferred format. Submit up to 10 URLs, get back structured content. Per-URL failures appear in errors[] and do not fail the entire request.
- Read
List fetch usage
List fetch usage records for the authenticated user. Returns paginated results with URL, format, status, and metadata for each fetch operation.
Search
- Read
Search the web
Search the web and get structured results. Returns ranked results with titles, snippets, and URLs. Location and language resolution: - If location is set but language is not, the language auto-resolves to the most predominantly used language in that country.
- Read
List search usage
List search usage records for the authenticated user. Returns paginated results with query details, status, and result counts.
Browser Context Profiles
- Read
List Browser Context Profiles
List Browser Context Profiles
- Read
Get a Browser Context Profile
Get a Browser Context Profile
- Write
Create a Browser Context Profile
Create a Browser Context Profile
- Write
Update a Browser Context Profile
Update a Browser Context Profile
- Write
Create a profile setup session
Starts a temporary setup browser for a Browser Context Profile. The response returns sessionid, cdpurl, baseurl, timeoutseconds, and expiresat. Connect your Playwright, Puppeteer, or CDP client to cdpurl, log in, then save the setup session with sessionid.
- Write
Save a profile setup session
Captures cookies, local storage, and session storage from a running setup session into the Browser Context Profile. Requires the sessionid returned by setup-session.
- Write
Upload Browser Context Profile state
Upload Browser Context Profile state
- Destructive
Delete a Browser Context Profile
Delete a Browser Context Profile
- Destructive
Cancel a profile setup session
Cancels a running Browser Context Profile setup session and discards unsaved setup state. Requires the sessionid returned by setup-session.
- Destructive
Delete a saved profile domain
Delete a saved profile domain
Vault
- Read
List vault connections
List all connected vault providers for the authenticated user.
- Read
List vault items
List all vault items currently available from connected providers.
- Write
Connect a vault provider
Connect a supported password manager and immediately sync display-safe credential metadata.
- Write
Sync vault items
Sync items from connected providers and return merged item state plus sync counters.
- Destructive
Disconnect a vault provider
Disconnect a vault provider and remove its stored enabled items.
Wallet
- Read
Get wallet
Get the caller's wallet: current balance, auto-reload configuration, per-product contract rates, and any in-flight top-up.
- Write
Top up wallet (agent-paid)
Add money to the wallet, paid by the calling agent over the Machine Payments Protocol (MPP). The first call answers 402 with a payment challenge in WWW-Authenticate; retry with a Payment-Authorization credential to settle the charge and credit the wallet.
Research
- Read
Attach to research progress
Reconnect with Last-Event-ID. Disconnecting does not cancel research. Retained events expire; poll the run for durable results.
- Read
List research runs
List saved research runs with optional filtering by status, query text, and date range. Cursors are only valid with the sort direction used to generate them.
- Read
Get research run by ID
Retrieve one saved research run, including the plan, evidence, and per-iteration working state.
- Write
Run research with SSE streaming
Search, evaluate sources, and synthesize a cited report from a natural-language query. Accepted requests always return a Server-Sent Events stream, including when stream is false.
- Write
Start research asynchronously
Enqueue research and return its ID after durable dispatch. Poll GET /v1/research-run/{id} or attach to its events endpoint.
- Destructive
Cancel research run by ID
Cancel a running research run. Idempotent: an already-terminal run returns 200 with cancelled false and its current status.
Monitor
- Read
List Monitors
GET /v1/monitors
- Read
Get a Monitor
GET /v1/monitors/{id}
- Write
Create a Monitor
Create a Page (fetch) or Topic (search) Monitor and capture its baseline run.
- Write
Update a Monitor
Send status alone to pause or resume, or schedulecron with optional fields to edit.
- Write
Run a Monitor now
Run a Monitor immediately, outside its schedule.
- Destructive
Delete a Monitor
Delete a Monitor and its schedule.
Datasets
- Read
List datasets
Newest first. Pass pagination.nextcursor as cursor for the next page.
- Read
Get a dataset
GET /v1/datasets/{id}
- Read
Get rows
One page of rows in insertion order, with the typed columns. Pass nextcursor as after for the next page.
- Read
Follow a run
Server-sent events until the run ends: status frames ({status, error?}; queued until a worker picks the run up, then running), row frames ({rowid, data, deleted?}) as cells land, log frames ({text}), then done.
- Write
Create a dataset
Describe the dataset in a prompt, or send a spec from /v1/datasets/plan.
- Write
Plan a dataset schema
The schema the agent would use for a prompt, to review or edit before creating.
- Write
Populate a dataset
Fills the dataset to its target from what the last run learned. reset empties it first.
- Destructive
Delete a dataset
Removes the dataset and its rows. Cancel a live run first.
- Destructive
Cancel a run
POST /v1/datasets/{id}/runs/{runId}/cancel
No operations match.
See How connections work for access and permissions, or browse all integrations.