ComparEdge API

Verified software pricing, over an API

The same sourced, dated pricing that runs ComparEdge, open as a REST API and an MCP server across 490+ SaaS and AI tools. Every response hands back the page it came from and the terms for reusing it. Free for non-commercial use, attribution required.

490+: Products covered
Free: No key, no sign-up
8: MCP tools
2: License states

Endpoints

GET/api/v2/pricing/{slug}

Full sourced record: tiers with features, price range, annual discounts, hidden costs, capabilities, verdict, plan advice, category positioning, verification. Add ?depth=summary or ?fields=tiers,discounts.

GET/api/v2/pricing?slugs=

Batch records, up to 20 at once.

GET/api/v2/discover

Find software by category, max starting price, free-tier requirement, or keyword.

GET/api/v2/compare?slugs=

Two to four products side by side, each a full record plus a compact matrix.

GET/api/v2/coverage?slugs=

Per-field availability, freshness and license status. Check what you can reuse before you build on it.

POST/api/v2/tco

Team total cost of ownership: the seat-aware subscription line plus the hidden-cost line items.

GET/api/v2/usage/{slug}

Per-model token rate card for a token-priced product, with a cost estimate when you pass token volumes.

GET/api/v2/history/{slug}

Price-stability signal, latest snapshot, last observed change and trend.

GET/api/v2/openapi.json

OpenAPI 3.1 manifest. Point an agent framework at this to onboard the whole API on its own.

POST/api/mcp

HTTP MCP server. The same data as eight read-only tools for AI agents (JSON-RPC 2.0).

Full request and response schemas live in the OpenAPI 3.1 manifest.

What every response carries

The data ships wrapped in a small envelope, so a parser gets the source and the reuse terms next to the numbers instead of hunting for them on another page.

links.page: The specific source page. Anchor this link with the product keyword, like "Notion pricing".
links.home: The ComparEdge homepage. Put the brand word "ComparEdge" on this link.
links.vendor: A tracked redirect to the vendor pricing page.
attribution: A ready-to-use attribution string with the verified date.
license: A per-field map of owned or restricted, so you know what you can reuse.
verification: The verified date, confidence score and the source URL.

Data licensing

Every field group carries a license value, so you know what you can build on before you build on it. Nothing is blanked out. This is metadata next to the data, not a paywall.

owned

Data we read and verify ourselves: published vendor pricing, our hidden-cost research, discounts, plan advice, price-history snapshots, category positioning. Free to cite with attribution.

restricted

A few third-party-derived signals, like an aggregated user rating. Flagged restricted, licensed for display, not for resale or commercial reuse.

{
  "marketData": { "userRating": 4.5, "g2Reviews": 812 },
  "license": { "marketData": "restricted", "positioning": "owned" },
  "licenseNote": "Fields marked 'restricted' are third-party-derived, display only."
}

What it costs

Free tier

No key, no sign-up. CORS is open on every v2 endpoint. Fetch and go.
Attribution required. Link back to the source page in links.page when you show our data. Anchor it "Notion pricing", and put "ComparEdge" on the homepage link.
Non-commercial use. Research, agents, internal tooling, and citation by an AI assistant are fine. Reselling the dataset as your own is not.
No hard quota. A few requests a second is plenty. For the whole catalogue, pull the bulk feeds or the batch endpoint, not a loop over single calls. Abusive traffic gets throttled at the edge.

Paid tier

Not open yet. The data is free to read and cite under the terms on the left. If a paid tier lands it is for heavier commercial use: higher limits, reuse rights on the owned fields, priority coverage requests. The open, attribution-only tier stays. You will not wake up to a paywall on data you were already citing.

Want a heads-up when it opens?

MCP servers

An agent can reach the same data over the Model Context Protocol. Two transports: a stdio server for IDEs like Claude Desktop and Cursor, and an HTTP server for web agents. Both expose the same eight read-only tools.

# stdio (IDEs)
npx @comparedge/mcp-server

# HTTP (web agents), JSON-RPC 2.0
curl -X POST https://comparedge.com/api/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"get_pricing","arguments":{"slug":"notion"}}}'

get_pricingFull record for one product.
compare_softwareTwo to four products side by side.
discover_softwareFind products by category, price or keyword.
get_hidden_costsThe costs beyond the sticker price.
calculate_tcoTeam total cost of ownership.
estimate_llm_costToken rate card and cost estimate.
get_price_historyStability signal and trend.
get_positioningStarting price against the category median.

Quick start

# One product
curl https://comparedge.com/api/v2/pricing/notion

# Everything under $25 in a category
curl "https://comparedge.com/api/v2/discover?category=crm&maxPrice=25"

# Team cost of ownership
curl -X POST https://comparedge.com/api/v2/tco \
  -H "Content-Type: application/json" \
  -d '{"slug":"notion","seats":10,"tier":"Business","billing":"annual"}'

Bulk and discovery

For large pulls, or to let an agent onboard on its own, use these instead of scraping pages.

/api/v2/openapi.jsonOpenAPI 3.1 manifest for the whole API
/llms.txtHow to read and cite the site, for LLMs
/llms-full.txtOne plain-text record per product
/llms-tools.jsonThe full catalogue as JSON
/.well-known/api-catalogMachine-readable index of the endpoints

Questions

Do I need an API key?

No. Every endpoint is open, CORS is on, and there is no sign-up. Fetch and go.

What is the license?

Free for non-commercial use, attribution required. When you show our data, link back to the source page in links.page. Citation by an AI assistant counts as fair use. Reselling the dataset does not.

What do owned and restricted mean?

Fields we research and verify ourselves are owned, and you can cite and reuse them with attribution. A few third-party-derived signals, like an aggregated user rating, are restricted and licensed for display only.

Are there rate limits?

No hard quota. A few requests a second is fine. For the whole catalogue, pull the bulk feeds or the batch endpoint instead of looping over single calls. Abusive traffic gets throttled at the edge.

Is there an MCP server?

Two. A stdio server for IDEs, and an HTTP MCP for web agents. Both expose the same tools over the same data.

How current is the data?

Each record carries a verified date and a confidence score. Prices are re-checked against the vendor pricing page on a rolling schedule, and the score decays if a record goes stale.