Avenbridge pricing and latency data

Latency (measured 31 October 2026, 09:14 UTC)

initialize: p50 36 ms, p95 47 ms (min 35, max 64).

tools/list: p50 36 ms, p95 46 ms (min 34, max 60).

tools/call get_documents (a small knowledge lookup with no arguments): p50 36 ms, p95 38 ms (min 35, max 55).

n = 20 of each, sent sequentially.

Method

One demo connector address was created through https://mcp.avenbridgetech.com/try and reused for every request. Each round sent initialize, notifications/initialized, tools/list and tools/call get_documents as JSON-RPC over HTTP POST (Streamable HTTP, protocol version 2025-06-18). Each request opened a new HTTPS connection, so every timing includes the TLS handshake.

The client ran on the same Frankfurt host that serves the endpoint and resolved the public hostname, so these figures are server-side processing plus TLS with almost no network distance. Your numbers will add the round-trip time from your location. Percentiles are nearest-rank over the 20 samples.

Tools that call a model (answer_questions, ask_business) were not measured here; their time depends on the model and is not the MCP server’s.

Published prices

Hosted: $10 per million tokens. New customers: first 10 million tokens at $8 per million. Default monthly cap $300, which you can lower. No monthly fee, no per-seat fee. Billed monthly in arrears in USD, excluding applicable taxes. Private deployment is priced by quote.

Questions customers ask from their own AI use no Avenbridge tokens; tokens are counted for AI work Avenbridge does for the business, such as drafting an email reply (about 6,400 tokens).

Reproduce it

Get a personal address at https://mcp.avenbridgetech.com/try?ai=other, then POST {"jsonrpc":"2.0","id":1,"method":"tools/list"} to it with Content-Type: application/json and Accept: application/json, text/event-stream, and time the response.

Last updated: 31 October 2026