# Parallel

Web search API that takes an objective and keyword queries and returns ranked pages with LLM-oriented excerpts.

- Provider: Parallel Web Systems (https://parallel.ai)
- Category: Search the web
- Type: Hosted API
- Pricing model: Pay as you go, per request [1]
- Free tier: Up to 5,000 requests a month free [1]
- Authentication: API key in `x-api-key` header; OAuth or no key for the MCP server [8]
- Connect via: REST API, MCP server, Python and TypeScript SDKs [4]
- Read from the provider's pages on 2026-11-04
- Page: https://supermarket.dev/tools/search-the-web/parallel

## Pricing

### Billing
- Pricing model: Pay as you go, per request [1]
- Free tier: Up to 5,000 requests a month free [1]
- Card required for free tier: Not published
- Minimum spend: Not published
- Billing period: Not published
- Free credits: Up to $80 at signup plus $5 a month [1]
- Free MCP access without a key: `web_search` is free at lower rate limits, no key [8]

### Prices
From the docs pricing page: `turbo` and `fast` $1, `basic` and `advanced` $5 per 1,000 requests with 10 results; extra results $1 per 1,000. No conversion needed. Requests default to `advanced`, so the default price is $5.
| Mode | $ per 1,000 requests | Typical latency |
|---|---|---|
| `advanced` (default) | 5 | about 3 s |
| `basic` | 5 | about 1 s |
| `fast` | 1 | about 700 ms |
| `turbo` | 1 | about 200 ms |
| Each additional result beyond 10 | 1 | - |
Source [2]

### Plans
Other endpoints listed as extras.
| Item | Price |
|---|---|
| Image Search, `fast` | $1 per 1,000 requests |
| Image Search, `advanced` | $5 per 1,000 requests |
| Extract | $1 per 1,000 URLs |
| Task API | $5 to $2,400 per 1,000 runs |
Source [2]

### Other charges
- Spend limit: Monthly, set by organization admins; notify-only [9]
- MCP usage billing: Calls with OAuth or an API key are billed at the API rates [8]

## Tools

### MCP server
- Official MCP server: Yes, remote [8]
- Address or package: `https://mcp.parallel.ai/mcp` [8]
- tools/list count without credentials: 3 [15]

### MCP tools
| Tool | What it does | Read-only |
|---|---|---|
| `web_search` | Searches the web and returns excerpts (uses `fast` mode) | Yes |
| `web_fetch` | Reads specific URLs, returns relevant parts or full page as markdown | Yes |
| `web_decisions` | Answers typed questions from current web evidence | Yes |
| `web_research` | Researches one question and returns a cited answer; needs sign-in or API key | Not stated (not listed without credentials) |
Source [8]

### API
| Endpoint | What it does |
|---|---|
| `POST /v1/search` | Search: pages with excerpts for an objective and queries |
| `POST /v1/images/search` | Image Search |
| `POST /v1/extract` | Extract: page content as markdown (extra endpoint) |
| `POST /v1/decisions` | Decisions: typed answers grounded in web evidence (extra endpoint) |
| `POST /v1/tasks/runs` | Task API: deep research and enrichment (extra endpoint) |
| `POST /v1/responses` | Responses API, OpenAI-compatible (extra endpoint) |
Source [3]

### Features
- Results per query: 10 by default; up to 20 [6]
- Content returned: Title, URL, publish date and markdown excerpts [12]
- Search types: Web search in four modes: `turbo`, `fast`, `basic`, `advanced`; image search is a separate endpoint [5]
- Freshness controls: `after_date` publication filter; `fetch_policy` chooses indexed or live content [7]
- Domain filters: `include_domains`, `exclude_domains` (domains or path prefixes), max 200 entries [7]
- Storing results: Customer terms bar copying or caching significant parts of output to build data-selling services [11]
- Location: Country code via `advanced_settings.location` [6]

## Authentication

### Methods
| Method | Used for | How |
|---|---|---|
| API key | REST API and MCP server | Send `x-api-key: $PARALLEL_API_KEY`; MCP also accepts `Authorization: Bearer` |
| OAuth sign-in | MCP server, `web_research` and higher limits | Connect to `https://mcp.parallel.ai/mcp?login` |
| No key | MCP `web_search`, `web_fetch`, `web_decisions` at lower rate limits | Connect to `https://mcp.parallel.ai/mcp` |
Source [8]

### Getting access
- Who creates the account: A person signs up on Platform; the MCP server works without an account [9]
- Where to get a key: `platform.parallel.ai` [4]
- Other requirements: Not published
- Agent account API: A Parallel Account API uses device-based OAuth 2.0 [13]

## Setup

### Addresses
- REST API: `https://api.parallel.ai/v1/search` [4]
- MCP server: `https://mcp.parallel.ai/mcp` [8]
- OpenAPI spec: `https://docs.parallel.ai/public-openapi.json` [13]
- Docs for agents: `https://docs.parallel.ai/llms.txt` [13]
- Status page: `https://status.parallel.ai` [13]

### Install
| What | Command |
|---|---|
| MCP, Claude Code | `claude mcp add --transport http parallel https://mcp.parallel.ai/mcp` |
| MCP, Codex | `codex mcp add parallel --url https://mcp.parallel.ai/mcp` |
| MCP, Gemini CLI | `gemini extensions install https://github.com/parallel-web/search-mcp` |
| Python SDK | `pip install parallel-web` |
| TypeScript SDK | `npm install parallel-web` |
Source [8]

## Limits

### Rate limits
- Search: 600 requests per minute [3]
- Higher limits: Request from support [3]
- MCP without a key: Free at lower rate limits; number not stated [8]

### Other limits
- Max results per request: 20 [6]
- Source filter entries: 200 per request, include and exclude combined [7]
- Turbo languages: English and Japanese [5]
- EU endpoint: `basic` mode only [5]
- URLs per `web_fetch` call: 20 [8]

## Privacy

### Your data
- Data location: US data centers by default; EU endpoint under enterprise agreement [9]
- Retention: Not published
- Retention (EU endpoint): Request and response content not retained [10]
- Used for training: Yes under standard terms, not linked to any individual; excluded on EU endpoint and zero-retention plans [11]
- Zero data retention: Available on enterprise plans [9]

### Compliance
- Certifications: SOC 2 Type I and Type II [9]
- HIPAA: Not published

## Sources

1. Pricing (website): https://parallel.ai/pricing
2. API pricing (docs): https://docs.parallel.ai/getting-started/pricing.md
3. API rate limits: https://docs.parallel.ai/getting-started/rate-limits.md
4. Search API quickstart: https://docs.parallel.ai/search/search-quickstart.md
5. Search modes: https://docs.parallel.ai/search/modes.md
6. Advanced search settings: https://docs.parallel.ai/search/advanced-search-settings.md
7. Source policy: https://docs.parallel.ai/search/source-policy.md
8. Parallel MCP: https://docs.parallel.ai/search/search-mcp.md
9. FAQs: https://docs.parallel.ai/resources/faqs.md
10. EU data residency: https://docs.parallel.ai/resources/eu-data-residency.md
11. Customer terms: https://www.parallel.ai/customer-terms
12. OpenAPI spec: https://docs.parallel.ai/public-openapi.json
13. llms.txt: https://docs.parallel.ai/llms.txt
14. Trust center: https://trust.parallel.ai
15. Checked by supermarket.dev