Last Price
ai.last-price/pricingVersion 0.10.0Compute and verify prices from an agent, metered per call. Works keyless on a demo tenant.
Status
Responds253 msAnswered our handshake without sign-in.- Last checked
- Nov 14, 2026, 13:31 UTC
- Latency
- 253 ms
- Remotes responding
- 2 of 2
- Sign-in
- No sign-in
- Hosting
- Hosted outside the EU (detected)
- Registry data as of
- Nov 14, 2026, 22:14 UTC
Data from the official MCP Registry, checked by run.zero. How we check
Connect
Streamable HTTP
https://api.last-price.ai/api/mcp/rpcResponds276 msNo sign-in
Claude Code
claude mcp add --transport http pricing https://api.last-price.ai/api/mcp/rpcClaude Desktop and claude.ai
Settings > Connectors > Add custom connector
Name: pricing
Remote MCP server URL: https://api.last-price.ai/api/mcp/rpcCursor
{
"mcpServers": {
"pricing": {
"url": "https://api.last-price.ai/api/mcp/rpc"
}
}
}VS Code
code --add-mcp '{"name":"pricing","type":"http","url":"https://api.last-price.ai/api/mcp/rpc"}'Streamable HTTP
https://api.last-price.ai/api/mcp/rpc?profile=leanResponds253 msNo sign-in
Claude Code
claude mcp add --transport http pricing 'https://api.last-price.ai/api/mcp/rpc?profile=lean'Claude Desktop and claude.ai
Settings > Connectors > Add custom connector
Name: pricing
Remote MCP server URL: https://api.last-price.ai/api/mcp/rpc?profile=leanCursor
{
"mcpServers": {
"pricing": {
"url": "https://api.last-price.ai/api/mcp/rpc?profile=lean"
}
}
}VS Code
code --add-mcp '{"name":"pricing","type":"http","url":"https://api.last-price.ai/api/mcp/rpc?profile=lean"}'Built your own server? run.zero hosts MCP servers on European infrastructure.
Tools
10 tools
Read-only | Idempotent | Open world | ||
|---|---|---|---|---|
call_toolRun any Last Price tool by name, including one this connection does not list. Find the name and its arguments with search_tools first. What it does, and whether it changes anything, is what that tool says. | Yes | No | No | Yes |
compute_priceCompute a price. Returns the price, the function that produced it, the metered cost and a `trace` of each stage. Send currency plus unit_cost (what the work costs you) or current_price. Missing inputs return an error naming them. For a competitor offer, put competitor_price, competitor_currency and competitive_action in metadata. To pay for this one call with x402, pass payment (call without it to read the price). A result with valid_until (RFC 3339) holds only until then. confidence is the function's own 0 to 1 score for its price, not a probability that a buyer accepts. price_statement says what the answer is: a third party's estimate for you, or the workspace's commitment when a signed quote is attached. | No | No | No | Yes |
get_receiptGet the receipt for one of your workspace's priced calls, by its usage event id (the id list_usage_events returns and report_outcome takes): what was priced, by which function and agent, when, the amount, how it was paid, and the outcome if one was reported. Receipts are made only when you ask, never sent for every call. For CSV or a printable page, call GET /api/receipts/{id} with format=csv or format=html. Needs a Last Price credential with price compute or billing read scope. Needs a claimed workspace: a sandbox key from self-registration is refused (403) until a person claims the workspace. | No | Yes | Yes | No |
list_functionsList the pricing functions registered for the authenticated tenant. | No | Yes | Yes | No |
pricing_conversationAsk Last Price what pricing inputs are missing. Start with request: {}, return its questions to the caller, then send continuation.request and answers. Prepare is free and does not compute. When ready, action compute uses the existing authenticated pricing API; configure a credential first. Never invent answers or treat a continuation as authorization. | No | No | No | No |
register_agentSign up with no key and no human: creates a sandbox workspace of your own and returns its API key (shown once), its daily limits and a claim link a person can open to take the workspace into their account. Pass name (2 to 64 characters, such as your-system/role-name), and optionally contact_email and intended_use. On the hosted MCP server this session goes on using the new key, so the tools that need a credential work at once with no header to add; elsewhere send the returned key as the x-api-key header. Public: needs no credential. | No | No | No | No |
report_outcomeReport whether a price from compute_price was used: once the buyer has accepted or turned down a price, report it. Pass receipt.id from that result as usage_event_id, or omit it to report the last price compute_price returned on this connection or tools instance, once (the hosted MCP endpoint remembers nothing, so pass it there). Outcomes are what later routing and confidence learn from. With outcome pricing on, a rejection within the call's outcome window gives back part of what it cost; the result says what was refunded. Needs a Last Price credential. Report only what actually happened. Needs a claimed workspace: a sandbox key from self-registration is refused (403) until a person claims the workspace. | No | No | No | No |
search_toolsFind the Last Price tool for a task. This connection lists only the tools that reach a price; the server has over 300 more (pricing functions, routing, experiments, webhooks, billing, orders, the marketplace and more). Pass query, a few words for the job, and get back the best matching tools with their full input schemas. Then call one with call_tool, or by its name. With no query it lists every tool name. | No | Yes | Yes | No |
use_api_keyUse a Last Price API key you already hold (a saved file, LASTPRICE_API_KEY) in this MCP connection, when it was opened without one: pass api_key. On the hosted MCP server this session then runs as that key's agent, so the tools that need a credential work at once with no header to add; it lasts until the session ends or the key is revoked or rotated. The key is checked, never echoed back. A connection that already has a credential keeps it. Public: needs no other credential. | No | No | Yes | No |
verify_priceAsk whether a price makes sense for its context, rather than what the price should be. Returns a verdict (within_band, too_high, too_low, or unknown), the band it was judged against, a signed deviation from the reference, a confidence, and the evidence behind it. A verdict of "unknown" means the price could not be judged, which is not the same as within_band. Act on next_step: buy (within band on confident, independent evidence), negotiate (too high), walk_away (far too high, confidently) or check_more (unknown, too low, or weak evidence); next_step_reason says why. Pass quote_valid_until from the compute_price result the price came from, and where quote expiry is switched on the answer says quote_expired when that quote is stale. Where paid verdicts are switched on, a caller with no key can pay for one verdict with x402: pass payment, a signed payment for the price and terms listed under payments.x402.verify_price in /.well-known/agents.json; a verdict of unknown is not charged. Where buyers' spend limits are switched on and metadata.category names a category, the evidence includes buyer_budget: what buyers' own spend checks allowed for it (buyer_envelope), reported and never counted in the verdict. price_statement says what the answer is: a third party's estimate whose band is an acceptance range around the reference, not an interval of value. | No | No | No | Yes |
Details
- Registry status
- Active
- Published
- Oct 29, 2026
- Updated
- Oct 29, 2026
- Repository
- Not given
- Categories (suggested)
- Invoicing and accountingSuggested categories are assigned automatically from the registry text and may be wrong. Verified operators can correct them.
- GDPR
- Not declared