Measured 5 October 2026
OpenAPI and Swagger specifications to MCP tools
How many public OpenAPI specifications convert into MCP tools?
Gatana converted 299 of 300 randomly selected public OpenAPI and Swagger specifications sampled from apis.guru into MCP tools (99.7%), and every operation became a tool. The one failure is a webhook specification with no paths. The median input schema per tool is 1,424 bytes. Measured on 5 October 2026.
| First run | After the fix | |
|---|---|---|
| Specifications converted | 262 of 300 (87.3%) | 299 of 300 (99.7%) |
| Swagger 2.0 | 121 of 122 | 122 of 122 |
| OpenAPI 3.x | 141 of 178 | 177 of 178 |
| Tools generated | 6,964 | 10,834 |
| Largest specification | NetBox, 386 tools | GitHub REST API, 845 tools |
| Conversion time per specification, median (maximum) | 0.2 s (8 s) | 0.15 s (3.7 s) |
| Input schema per tool, median | 11,276 bytes | 1,424 bytes |
| Input schema per tool, 90th percentile | 94,668 bytes | 4,808 bytes |
| Tool names over 64 characters | not counted | 84 of 10,834 (0.8%) |
How we measured it
A sample of 300 specifications from apis.guru (122 Swagger 2.0, 178 OpenAPI 3.x, seed 20261005), each created as an OpenAPI server in one Gatana organization through the public API, the tools refreshed and read back. The first run converted 262. All but one of the 38 failures were defects in our converter, fixed the same day; the table shows both runs.
Full write-up: We Converted 300 Public OpenAPI Specs to MCP Tools. Here Is What Broke.
Raw data: CSV, 300 rows: the sample, both result sets, schema sizes and the error texts
Measured 7 October 2026
Latency added by the gateway
How much latency does the Gatana gateway add to a tool call?
Gatana added 6.6 ms per tool call in our test on 7 October 2026, after a release that keeps the caller lookups in the gateway process for 30 seconds, with the gateway and the upstream server on MCP revision 2026-07-28. Against a 2025-era upstream the same build added 9.6 ms. Measured on a Hetzner machine with an AMD Ryzen™ 5 3600, warm sequential calls inside one cluster.
| Measurement | Run 1 | Run 2 | Run 3 |
|---|---|---|---|
| Direct echo, public hostname, median | 1.5 ms | 1.4 ms | 1.5 ms |
| Through the gateway to echo, 2026-07-28 upstream, median | 7.9 ms | 8.3 ms | 8.1 ms |
| Through the gateway to echo, 2026-07-28 upstream, 95th percentile | 26.1 ms | 21.0 ms | 46.0 ms |
| Through the gateway to echo, 2026-07-28 upstream, 99th percentile | 38.2 ms | 38.5 ms | 65.5 ms |
| Gateway median minus direct median | 6.4 ms | 6.9 ms | 6.6 ms |
| Through the gateway to echo, 2025-era upstream, median | 12.2 ms | 10.9 ms | 10.5 ms |
| Successful built-in call with no upstream, median | 6.9 ms | 7.1 ms | 7.5 ms |
How we measured it
21,900 successful timed calls from a probe inside the same cluster as the gateway and echo MCP server, hosted on Hetzner with an AMD Ryzen™ 5 3600 processor. Three interleaved repetitions used kept-alive connections and 20 warm-up calls per run. Every gateway path ran with a 2025-era client and with a 2026-07-28 client, against an upstream recorded as 2025-era and against the same echo server recorded as 2026-07-28. The estimate is the median of the three differences between the gateway median and the direct public median of the same client era: 6.4, 6.9 and 6.6 ms for the all-2026 path, 10.7, 9.6 and 9.2 ms with a 2025-era upstream. It is not a percentile of individual request overhead or a concurrent-load test.
Full write-up: How Much Latency Does an MCP Gateway Add? 6.6 Milliseconds in Our Latest Test.
Raw data: JSON: all 27 runs, percentiles, hardware, both protocol eras, and the overhead calculation
Measured 7 October 2026, superseded by a later run the same day
Latency added by the gateway
How much latency does the Gatana gateway add to a tool call?
Gatana added 9.5 ms per tool call in our test on 7 October 2026, with the gateway and the upstream server on MCP revision 2026-07-28, which has no per-call handshake. Against a 2025-era upstream the same build added 13.5 ms. Measured on a Hetzner machine with an AMD Ryzen™ 5 3600, warm sequential calls inside one cluster.
| Measurement | Run 1 | Run 2 | Run 3 |
|---|---|---|---|
| Direct echo, public hostname, median | 1.4 ms | 1.4 ms | 1.5 ms |
| Through the gateway to echo, 2026-07-28 upstream, median | 10.9 ms | 10.7 ms | 11.1 ms |
| Through the gateway to echo, 2026-07-28 upstream, 95th percentile | 22.7 ms | 21.6 ms | 23.4 ms |
| Through the gateway to echo, 2026-07-28 upstream, 99th percentile | 35.8 ms | 30.8 ms | 55.0 ms |
| Gateway median minus direct median | 9.5 ms | 9.3 ms | 9.6 ms |
| Through the gateway to echo, 2025-era upstream, median | 14.9 ms | 14.3 ms | 15.0 ms |
| Successful built-in call with no upstream, median | 10.0 ms | 11.0 ms | 10.4 ms |
How we measured it
21,900 successful timed calls from a probe inside the same cluster as the gateway and echo MCP server, hosted on Hetzner with an AMD Ryzen™ 5 3600 processor. Three interleaved repetitions used kept-alive connections and 20 warm-up calls per run. Every gateway path ran with a 2025-era client and with a 2026-07-28 client, against an upstream recorded as 2025-era and against the same echo server recorded as 2026-07-28. The estimate is the median of the three differences between the gateway median and the direct public median of the same client era: 9.5, 9.3 and 9.6 ms for the all-2026 path, 13.5, 13.1 and 13.6 ms with a 2025-era upstream. It is not a percentile of individual request overhead or a concurrent-load test.
Full write-up: How Much Latency Does an MCP Gateway Add? 9.5 Milliseconds in Our Latest Test.
Raw data: JSON: all 27 runs, percentiles, hardware, both protocol eras, and the overhead calculation
Measured 7 October 2026, superseded by a later run the same day
Latency added by the gateway
How much latency does the Gatana gateway add to a tool call?
Gatana added about 14 ms per tool call in our test on 7 October 2026. Gatana was deployed on a Hetzner dedicated machine with an AMD Ryzen™ 5 3600 processor. Across three paired runs, the added latency estimate was 13.4 ms. The gateway path had a 14.8 ms median against 1.4 ms direct. These are warm sequential calls inside one cluster.
| Measurement | Run 1 | Run 2 | Run 3 |
|---|---|---|---|
| Direct echo, public hostname, median | 1.4 ms | 1.5 ms | 1.4 ms |
| Through the gateway to echo, median | 15.8 ms | 14.4 ms | 14.8 ms |
| Through the gateway to echo, 95th percentile | 44.4 ms | 33.4 ms | 35.9 ms |
| Through the gateway to echo, 99th percentile | 96.6 ms | 96.2 ms | 73.9 ms |
| Gateway median minus direct median | 14.4 ms | 12.9 ms | 13.4 ms |
| Successful built-in call with no upstream, median | 10.8 ms | 9.7 ms | 9.8 ms |
How we measured it
8,400 successful timed calls from a probe inside the same cluster as the gateway and echo MCP server, hosted on Hetzner with an AMD Ryzen™ 5 3600 processor. Three interleaved repetitions used kept-alive connections and 20 warm-up calls per run. Each repetition had 1,000 direct public echo calls, 1,000 gateway echo calls, 500 successful skills_list_skills calls, and 300 direct internal echo calls. The overhead estimate is the median of the three differences between gateway and direct public path medians: 14.4, 12.9, and 13.4 ms. It is not a percentile of individual request overhead or a concurrent-load test.
Full write-up: How Much Latency Does an MCP Gateway Add? About 14 Milliseconds in Our Latest Test.
Raw data: JSON: all 12 runs, percentiles, hardware, and the overhead calculation
Measured 5 October 2026, superseded by the run of 7 October 2026
Latency added by the gateway
How much latency does the Gatana gateway add to a tool call?
A tool call through the Gatana gateway took 30.3 ms at the median against 1.4 ms for the same call made directly, so the gateway adds about 29 ms for authentication, authorization, tool resolution and the audit record. A real tool call takes 1.0 s at the median, so the overhead is about 3%. Measured on 5 October 2026.
| Path | Morning, median (95th pct.) | Afternoon after the change, median (95th pct.) |
|---|---|---|
| Direct to the server, public hostname | 1.2 to 1.4 ms (1.7 to 1.8 ms) | 1.2 to 1.4 ms (1.6 to 1.7 ms) |
| Through the gateway to the same server | 53.4 to 53.8 ms (64.8 to 66.9 ms) | 30.3 to 31.2 ms (37.4 to 41.9 ms) |
| Through the gateway, built-in tool with no upstream server | 45.6 ms (57.0 ms) | 21.9 ms (27.3 ms) |
| Gateway overhead per call | about 52 ms | about 29 ms |
| Share of a real tool call (1.0 s median, 2.8 s 90th pct.) | 5% at the median | 3% at the median, 1% at the 90th percentile |
How we measured it
10,800 tool calls from a probe pod to an echo MCP server in the same cluster as a gateway, over one kept-alive connection after 20 warm-up calls, direct and through the gateway. Measured in the morning and again after a change to the gateway's per-call path the same afternoon. The real-call baseline is the audit log of one customer organization, used with its consent.
Full write-up: How Much Latency Does an MCP Gateway Add? We Measured Ours: 52 Milliseconds, Then 29.
Raw data: JSON: every run with its percentiles, morning and afternoon
Measured 5 October 2026
Code mode against direct tool calls
How many tokens does code mode save against direct tool calls?
On two tasks with more data than a model should hold in context, code mode used 4 to 44 times fewer prompt tokens than direct tool calls (30,458 against 1.33 million on the largest task) and 17 to 51 times less wall time. At Claude Opus 5.5 prices the largest task saves 7.73 USD per run. The data stays in the sandbox; only the result reaches the model. Measured on 5 October 2026 with DeepSeek-V4.1-Flash.
| Task | Mode | Correct | Prompt tokens (median) | Seconds (median) | USD per run, DeepSeek | USD per run, Claude Opus 5.5 prices |
|---|---|---|---|---|---|---|
| Revenue by region, 10,000 orders | Direct | 2 of 5 | 1,330,460 | 617 | 0.389 | 7.88 |
| Code mode | 5 of 5 | 30,458 | 12 | 0.003 | 0.15 | |
| Word counts, 40 documents | Direct | 0 of 5 | 242,630 | 318 | 0.103 | 2.32 |
| Code mode | 5 of 5 | 59,153 | 19 | 0.006 | 0.27 | |
| Five largest of 500 files | Direct | 5 of 5 | 31,110 | 6 | 0.006 | 0.14 |
| Code mode | 5 of 5 | 17,670 | 9 | 0.003 | 0.09 |
How we measured it
A synthetic server with four data tools (10,000 orders, 500 files, 40 documents of 65,000 words) and three tasks with answers known in advance. DeepSeek-V4.1-Flash at temperature 0 ran each task five times with the four tools exposed directly and five times through Gatana's two code mode tools, 30 runs in all. Correctness was checked against ground truth computed outside the model. The DeepSeek cost is at DeepSeek list prices. The Claude Opus 5.5 cost prices the median prompt and completion tokens of each task at list prices, 4 USD per million input tokens and 20 USD per million output tokens, with no cache discount.
Full write-up: Code Mode vs. Direct Tool Calls: The Same Three Tasks, Five Times Each
Raw data: JSON: all 30 runs with tokens, tool calls, time and the answer given
How we measure
- Every figure has a date. Software changes, and a number without a date is a claim. The gateway latency figure changed within one day.
- The raw data is published next to the figure, as CSV or JSON, so anyone can check the arithmetic or repeat the run.
- We measure Gatana, not competitors. The direct-call figures are the baseline, not a rival product.
- Customer data appears only as aggregates, with the organization's consent, never per record.
- A repeated run does not overwrite the old one. The old figure stays on this page, marked as superseded, with the date of the new run.
Reading this with a script or an agent?
The same figures are published as Markdown at gatana.ai/measurements.md.