Pricing, limits, and integration
Access costs nothing extra at any tier: the REST API and MCP server are included on every plan, and the free Explore tier gets both at entry-level limits. That makes PromptWatch one of the only GEO tools where an agent stack can be built and tested before any money changes hands. Keys are generated self-serve under Settings > API Keys and passed as X-API-Key (REST) or Authorization: Bearer (MCP).
Rate limits are published rather than discovered in production, across three windows: an hourly quota, a per-minute burst and a per-second burst, shared per organization between the REST API and the MCP server. The documented ladder runs from about 150 requests a minute on entry plans to 334 on Business and 834 on Enterprise, with a separate universal cap of 1,000 requests a minute per IP. Note the docs' plan names lag the pricing page (Solo and Startup versus today's Essential and Professional), so confirm your tier's exact numbers in-app.
The practical agent pattern: authenticate Claude over OAuth for interactive work (querying visibility trends, competitor movements and citation sources conversationally), give automation a scoped API key for scheduled jobs (syncing prompt sets, pulling time series into your warehouse, updating action items), and let the write tools handle prompt lifecycle management so your tracking setup stays in code. The recipes and skills pages in the MCP docs sketch several of these workflows ready-made.
How much does the PromptWatch API actually cost?
PromptWatch publishes an API, but per-tier gating wasn't extracted automatically.
- Most likely scenario: API access is either included in all paid plans, or it's gated to higher tiers and we need to verify on the live docs page.
- Until verified, treat the question "which plan unlocks the API" as open. Don't budget on the assumption it's free with the cheapest paid plan.
API at a glance
Which AI clients can read PromptWatch data?
Not every AI assistant supports MCP natively. Here's the per-client picture for PromptWatch specifically.
| Client | Support | Notes |
|---|---|---|
| Claude (Desktop/Web) | Native MCP | OAuth browser-consent flow, no credentials to paste; or npx mcp-remote with a Bearer API key. |
| Claude Code | Native MCP | claude mcp add --transport http promptwatch https://server.promptwatch.com/mcp with an Authorization header, or the OAuth flow. |
| Cursor | Native MCP | Documented setup: server URL plus Authorization: Bearer YOUR_API_KEY header. |
| ChatGPT | API-only (custom code) | No documented ChatGPT MCP path; use the REST API from custom GPT actions or code. |
| n8n | API-only (custom code) | No native node found; the REST API with X-API-Key auth covers the same surface for workflow automation. |
PromptWatch also has an MCP server
If you're wiring PromptWatch into Claude / Cursor / Codex, the MCP server is usually less code than the REST API.
See PromptWatch MCP pageWhat you can build with the PromptWatch API
Three personas, three different shapes of build. Pick the row that matches how you actually ship.
Glue PromptWatch into your daily workflow
- Pull a digest of fresh activity into your morning Slack
- Trigger a one-off backfill when something looks off
- Pipe data into your own SQLite for ad-hoc queries
Sell PromptWatch reporting at scale
- White-label PromptWatch data into client-branded dashboards
- Run scheduled multi-account reports without logging into the PromptWatch UI
- Cross-reference with the rest of the client's stack (CRM, analytics, ad spend)
Production pipelines for PromptWatch
- Real-time sync into your warehouse on a cron
- Custom alerts the dashboard's built-in alerting can't express
- Backfill + idempotency handling for PromptWatch data older than the dashboard exposes
Quickstart template
# Starter template — endpoint paths and SDK package names come from # the PromptWatch API docs: https://promptwatch.com/docs curl -X GET "https://promptwatch.com/<your-endpoint>" \ -H "Authorization: Bearer $TOKEN" \ -H "Accept: application/json" | jq .
Template only. Endpoint paths, exact base URL, and auth header format come from promptwatch.com/docs.
Limits and gotchas
- OpenAPI is published — auto-generated clients work out of the box. Saves writing wrappers by hand.
- No official SDK detected. Plan on hitting the API directly with your HTTP client of choice.
- Rate limits: always read the docs page before scaling — the published limit is usually lower than the practical one and overages can be expensive.
- Structured JSON outputs — agent-friendly, no scraping required.
Agent-readiness verdict
A broad, well-documented API with an OpenAPI spec, Postman collection, published three-window rate limits and markdown docs, included on every plan down to the free tier. Minor doc drift on plan names and no labeled bulk-export endpoint keep it at 8 out of 10 rather than higher.
Best agent stack in the GEO category: official remote MCP server (server.promptwatch.com/mcp, streamable HTTP) with OAuth browser-consent auth or API key, 59 tools including write operations (create/update/delete prompts, monitors, brands, personas, generate content). REST API v2 with ~70 endpoints, OpenAPI spec, Postman collection, published three-window rate limits. MCP and API access on every plan including the free Explore tier. llms.txt + markdown docs + MCP recipes and skills guides.
PromptWatch API FAQ
Does PromptWatch have a public API?+
What auth does the PromptWatch API use?+
Is there an OpenAPI / Swagger spec I can auto-generate clients from?+
Does PromptWatch support webhooks?+
Should I use the API or the MCP server for Claude / Cursor / Codex?+
Sources
- PromptWatch official site: https://promptwatch.com
- API docs: https://promptwatch.com/docs
- OpenAPI spec: https://promptwatch.com/docs/v2/openapi.json
- MCP server: https://promptwatch.com/docs/mcp/introduction
10+ years in digital marketing. I review marketing software for AI-stack fit: real pricing, MCP and API support, and how cleanly each tool drops into an AI agent workflow, cross-checked against verified data and real user feedback.