Parallel integration
Parallel integration for AI agents.
Parallel integration for AI agents with secure authentication and server-side credential injection. Open Connector runs the OAuth, seals the token in an encrypted vault, and serves Parallel tools to your agent over MCP or a typed API — credentials injected server-side, every call audited, nothing leaving your infrastructure. Open source (AGPL-3.0) and self-hostable.
What your agents can do
Real Parallel actions, managed and audited.
Your user connects Parallel once; your agent can then web infrastructure for AI to search, extract, monitor, and reason over the world's information — scoped to the OAuth permissions you grant and the tool allowlist you configure. Every action is least-privilege and written to a tamper-evident audit trail.
- 1
Your user grants Parallel access once (OAuth) — the token lands in the vault.
- 2
Your agent calls a tool over MCP or the typed API; Open Connector injects the credential server-side.
- 3
Every routed call appends a hash-chained audit record — nothing leaves your infra.
Tools & triggers
Supported Parallel tools.
33 tools are generated from the published Parallel catalog. Descriptions are plain text; each action remains subject to its configured authentication and tool allowlist.
Showing 33 tools. All published catalog entries are included in this page's server-rendered HTML.
- Cancel FindAll Run
- Cancel a FindAll run.
- FindAll
- Cancel Monitor
- Cancel a monitor. Permanently stops the monitor from running. Cancellation is irreversible — create a new monitor to resume monitoring. Cancelling an already-cancelled monitor is a no-op.
- Monitor
- Chat Completions
- Chat completions. This endpoint can be used to get realtime chat completions. It can also be used with the Task API processors to get structured, research outputs via a chat interface.
- Chat API (Beta)
- Create Monitor
- Create a monitor. Monitors run on a fixed frequency to detect material changes in web content. Set `type=event_stream` to monitor a search query, or `type=snapshot` to monitor a specific task run's output. The monitor runs once immediately at creation, then continues on the configured schedule.
- Monitor
- Create Response
- Create a response. Generates an answer to the given input, grounded in live web research and annotated with URL citations. Set `model` to `parallel`; `reasoning.effort` (`low`/`medium`/`high`) controls how much research is performed, trading response time for answer quality. Returns an OpenAI-format `Response` as `application/json`, or a `text/event-stream` of OpenAI Responses SSE events when `stream=true`.
- Responses API
- Add Enrichment to FindAll Run
- Add an enrichment to a FindAll run.
- FindAll
- Extend FindAll Run
- Extend a FindAll run by adding additional matches to the current match limit.
- FindAll
- Extract
- Extracts relevant content from specific web URLs. The legacy Extract API reference (`/v1beta/extract` endpoint) is available [here](https://docs.parallel.ai/api-reference/legacy/extract-beta/extract), and migration guide is [here](https://docs.parallel.ai/extract/extract-migration-guide).
- Extract
- Fast Entity Search
- Return ranked entities matching a natural language objective. This endpoint performs a best-effort search optimized for low latency. To keep responses fast, it returns a fixed set of attributes and supports queries of limited complexity. For comprehensive match evaluation and enrichment, use the [FindAll API](https://docs.parallel.ai/findall-api/findall-quickstart).
- FindAll
- Retrieve FindAll Run Status
- Retrieve a FindAll run.
- FindAll
- Create FindAll Run
- Starts a FindAll run. This endpoint immediately returns a FindAll run object with status set to 'queued'. You can get the run result snapshot using the GET /v1beta/findall/runs/{findall_id}/result endpoint. You can track the progress of the run by: - Polling the status using the GET /v1beta/findall/runs/{findall_id} endpoint, - Subscribing to real-time updates via the /v1beta/findall/runs/{findall_id}/events endpoint, - Or specifying a webhook with relevant event types during run creation to receive notifications.
- FindAll
- Stream FindAll Events
- Stream events from a FindAll run. Args: request: The Shapi request findall_id: The FindAll run ID last_event_id: Optional event ID to resume from. timeout: Optional timeout in seconds. If None, keep connection alive as long as the run is going. If set, stop after specified duration.
- FindAll
- FindAll Run Result
- Retrieve the FindAll run result at the time of the request.
- FindAll
- Get FindAll Run Schema
- Get FindAll Run Schema
- FindAll
- Ingest FindAll Run
- Transforms a natural language search objective into a structured FindAll spec. The generated specification serves as a suggested starting point and can be further customized by the user.
- FindAll
- List Monitor Events
- List events for a monitor, newest first. Pass `event_group_id` to narrow results to a single execution. Otherwise returns all executions newest-first; use `next_cursor` to paginate. Set `include_completions=true` to also include no-change executions.
- Monitor
- List Monitors
- List monitors ordered by creation time, newest first. Monitors are sorted by `created_at` descending. `limit` defaults to 100. Use `next_cursor` from the response and pass it as `cursor` to fetch the next page. Pagination ends when `next_cursor` is absent. By default only `active` monitors are returned. Pass `status=cancelled` or both values to include cancelled monitors. The legacy Monitor API (`/v1alpha/monitors` endpoints) is documented under the `Monitor (Alpha)` tag.
- Monitor
- Retrieve Monitor
- Retrieve a monitor. Retrieves a specific monitor by `monitor_id`. Returns the monitor configuration including status, frequency, query, and webhook settings.
- Monitor
- Stream Task Run Events
- Streams events for a task run. Returns a stream of events showing progress updates and state changes for the task run. For task runs that did not have enable_events set to true during creation, the frequency of events will be reduced.
- Tasks
- Stream Task Run Events
- Streams events for a task run. Returns a stream of events showing progress updates and state changes for the task run. For task runs that did not have enable_events set to true during creation, the frequency of events will be reduced.
- Tasks
- Retrieve Task Run
- Retrieves run status by run_id. The run result is available from the `/result` endpoint.
- Tasks
- Retrieve Task Run Input
- Retrieves the input of a run by run_id.
- Tasks
- Create Task Run
- Initiates a task run. Returns immediately with a run object in status 'queued'. Beta features can be enabled by setting the 'parallel-beta' header.
- Tasks
- Retrieve Task Run Result
- Retrieves a run result by run_id, blocking until the run is completed.
- Tasks
- Stream Task Group Events
- Streams events from a TaskGroup: status updates and run completions. The connection will remain open for up to an hour as long as at least one run in the group is still active.
- Tasks
- Retrieve Task Group
- Retrieves aggregated status across runs in a TaskGroup.
- Tasks
- Create Task Group
- Initiates a TaskGroup to group and track multiple runs.
- Tasks
- Fetch Task Group Runs
- Retrieves task runs in a TaskGroup and optionally their inputs and outputs. All runs within a TaskGroup are returned as a stream. To get the inputs and/or outputs back in the stream, set the corresponding `include_input` and `include_output` parameters to `true`. The stream is resumable using the `event_id` as the cursor. To resume a stream, specify the `last_event_id` parameter with the `event_id` of the last event in the stream. The stream will resume from the next event after the `last_event_id`.
- Tasks
- Retrieve Task Group Run
- Retrieves run status by run_id. This endpoint is equivalent to fetching run status directly using the `retrieve()` method or the `tasks/runs` GET endpoint. The run result is available from the `/result` endpoint.
- Tasks
- Add Runs to Task Group
- Initiates multiple task runs within a TaskGroup.
- Tasks
- Trigger Monitor Run
- Trigger an immediate monitor run. Enqueues a one-off execution of the monitor outside its normal schedule. The monitor's regular schedule is not affected. An event is only emitted if the execution detects a material change. Cancelled monitors cannot be triggered.
- Monitor
- Update Monitor
- Update a monitor. Only fields explicitly included in the request body are changed. Pass `null` for `webhook` or `metadata` to clear those fields. Pass `type` and `settings` to update type-specific settings on an `event_stream` monitor. At least one field must be provided. Cancelled monitors cannot be updated.
- Monitor
- Search
- Searches the web. The legacy Search API reference (`/v1beta/search` endpoint) is available [here](https://docs.parallel.ai/api-reference/legacy/search-beta/search), and migration guide is [here](https://docs.parallel.ai/search/search-migration-guide).
- Search
FAQ
Parallel integration, answered
- How do AI agents use Parallel through Open Connector?
- Your user connects Parallel once with one of its cataloged authentication methods. Open Connector stores the credential in an encrypted vault and exposes Parallel tools to your agent over MCP or a typed API, with credentials injected server-side on each call.
- Is this a Parallel MCP server?
- Yes. Open Connector can serve Parallel as a named MCP server with a scoped allowlist and a per-user connection URL, so any MCP client can call Parallel actions with credentials injected server-side.
- Where do Parallel credentials live?
- In your own infrastructure. Open Connector keeps credentials in its own vault and injects them at call time, so they never leave your environment.
Give your agents Parallel — keep the keys.
Open source, self-hostable, with Parallel credentials that never leave your infrastructure. Run it from source today.