Find which of your MCP tools AI agents retry, ignore or never call — first-call success, empty answers, schema cost. Two lines. Never your arguments or results.https://getmcpulse.comhttps://getmcpulse.com/blog/live-on-product-hunthttps://getmcpulse.com/blog/live-on-product-huntSixteen metrics for MCP servers, two lines to install, and the per-client split that a server-wide average hides. Free for one server.Tue, 08 Sep 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/registry-strips-requiredhttps://getmcpulse.com/blog/registry-strips-requiredZero of 82,549 stored tools carry a top-level required array. 2,064 nested objects still carry theirs, which narrows the cause considerably.Fri, 04 Sep 2026 00:00:00 GMTEngineeringhttps://getmcpulse.com/blog/the-servers-that-get-it-righthttps://getmcpulse.com/blog/the-servers-that-get-it-right718 of 2,255 servers have no undescribed parameters, no colliding tools, and no unenforced closed sets. They are smaller, cheaper and more distinctive.Thu, 03 Sep 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/should-you-split-a-large-mcp-serverhttps://getmcpulse.com/blog/should-you-split-a-large-mcp-serverA decision framework with the corpus data as the input. Under fifteen tools the answer is no. Past sixty it is probably yes, and rewording will not save you.Wed, 02 Sep 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/mcp-tool-description-checklisthttps://getmcpulse.com/blog/mcp-tool-description-checklistNine checks, each with the rate it occurs across 4,749 public servers, so you can spend your time on the two faults that are actually likely.Tue, 01 Sep 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/why-the-model-wont-call-your-toolhttps://getmcpulse.com/blog/why-the-model-wont-call-your-toolSix causes, ordered by how often they turn up in a corpus of 4,749 servers. Start with the one that is 17.7% likely, not the one that feels most plausible.Sat, 29 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/how-to-test-your-tool-schemashttps://getmcpulse.com/blog/how-to-test-your-tool-schemasA twenty-minute pass — dump the real response, run four checks, fix the two faults that account for most of the corpus, then put one assertion in CI.Fri, 28 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/how-to-read-a-tools-listhttps://getmcpulse.com/blog/how-to-read-a-tools-listThe JSON your MCP server returns is the entire brief a model gets. Here is how to read it the way a model has to, field by field.Thu, 27 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/servers-that-publish-no-toolshttps://getmcpulse.com/blog/servers-that-publish-no-toolsThey are listed, installable and reachable, and tools/list comes back empty. This figure was 13 until we deduplicated the corpus — a lesson in counting.Wed, 26 Aug 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/json-schema-features-mcphttps://getmcpulse.com/blog/json-schema-features-mcpdefault appears on 17.2% of parameters — more than enum. format is 2.1%, pattern 1.6%, and examples 0.3%. Most of JSON Schema goes unused.Tue, 25 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/how-deep-mcp-schemas-nesthttps://getmcpulse.com/blog/how-deep-mcp-schemas-nest83.1% of tools are one level deep and the deepest is thirteen. Deeply nested schemas are better documented than flat ones, not worse — 18.2% against 21.8%.Sat, 22 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/should-you-prefix-tool-nameshttps://getmcpulse.com/blog/should-you-prefix-tool-names10.6% of servers do. They collide less, not more — 14.6% zero-distinctive against 18.5% — and the cost is bytes rather than clarity.Fri, 21 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/how-mcp-authors-name-toolshttps://getmcpulse.com/blog/how-mcp-authors-name-toolsOnly 33.4% of 82,549 tool names start with a common verb. get alone is 15.6%, and 12,892 tools begin with it — which is where the collisions live.Thu, 20 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/what-makes-a-good-descriptionhttps://getmcpulse.com/blog/what-makes-a-good-descriptionLength is not the answer. Across 75,914 tools the correlation between description length and distinctiveness is +0.034 — statistically nothing.Wed, 19 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/the-2530-tool-serverhttps://getmcpulse.com/blog/the-2530-tool-server866KB of schema, about 216,000 tokens on every connection, and every single tool carries a description. 34.3% of its parameters do not.Tue, 18 Aug 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/popular-vs-long-tailhttps://getmcpulse.com/blog/popular-vs-long-tailThe most-installed servers are bigger — 30.4 tools against 16.6 — and worse on both schema measures. 28.6% of their parameters are undescribed, against 20.8%.Sat, 15 Aug 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/valid-values-open-stringhttps://getmcpulse.com/blog/valid-values-open-string857 parameters across 4.4% of servers name their valid values in prose and leave the schema an open string. A one-line fix, and only one of the two is enforced.Fri, 14 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/enums-in-mcp-schemashttps://getmcpulse.com/blog/enums-in-mcp-schemas20,263 of 257,287 parameters constrain their values. 12.4% of string parameters do. One enum in the corpus has 3,517 values, and 270 have exactly one.Thu, 13 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/duplicate-tool-nameshttps://getmcpulse.com/blog/duplicate-tool-names9 servers out of 4,749 ship a duplicate tool name — 0.19%, and only 16 duplicated entries in the whole corpus. If this is on your review checklist, take it off.Wed, 12 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/what-mcp-servers-cannot-tell-youhttps://getmcpulse.com/blog/what-mcp-servers-cannot-tell-youYou ship a server, it gets listed, people install it. From that moment on you are working blind — and the directories report almost none of what you need.Tue, 11 Aug 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/which-descriptions-matterhttps://getmcpulse.com/blog/which-descriptions-matterOne failure scales sixty-three times with server size. The other is flat at every size. So fixing either does nothing for the other — they are separate jobs.Sat, 08 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/params-per-toolhttps://getmcpulse.com/blog/params-per-toolThe median MCP tool takes 2 parameters and 9.5% take none. The undescribed-parameter rate is worst on tools with exactly one — 26.9%, against 19.9% at four to six.Fri, 07 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/shared-tool-preambleshttps://getmcpulse.com/blog/shared-tool-preamblesOne server repeats the same 112 words at the start of all 193 of its tool descriptions. It was added to help. It is why its tools cannot be told apart.Thu, 06 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/why-models-pick-the-wrong-toolhttps://getmcpulse.com/blog/why-models-pick-the-wrong-tool17.6% of tools across 4,749 servers have no word setting them apart from a sibling. But 89.6% still have a distinctive name, which shrinks the claim and sharpens it.Wed, 05 Aug 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/first-call-successhttps://getmcpulse.com/blog/first-call-successUptime grades your infrastructure. Error rate grades your code. First-call success grades the part you wrote yourself — the words.Tue, 04 Aug 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/what-your-tool-list-costshttps://getmcpulse.com/blog/what-your-tool-list-costsThe median public MCP server spends about 1,250 tokens of schema before anyone asks a question. The 90th percentile spends 7,992. One spends 286,000.Sat, 01 Aug 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/how-many-toolshttps://getmcpulse.com/blog/how-many-toolsAcross 4,749 public servers, the share of tools with nothing to tell them apart rises from 0.5% to 32.4% as the tool count grows. That is a factor of sixty-three.Fri, 31 Jul 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/21-percent-of-parametershttps://getmcpulse.com/blog/21-percent-of-parameters55,383 of 257,287 parameters across 4,749 servers ship with nothing but a name and a type. Only 0.4% of tools do. That asymmetry is the whole finding.Thu, 30 Jul 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/reading-5000-mcp-schemashttps://getmcpulse.com/blog/reading-5000-mcp-schemasNot what models do with your tools — what they are handed first. 82,549 tools, 257,287 parameters, one tool in six with nothing to tell it from its neighbour.Wed, 29 Jul 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/empty-results-are-failureshttps://getmcpulse.com/blog/empty-results-are-failuresAn empty array is a 200. The model got nothing it could use, the user got a shrug, and no error was recorded anywhere. These are the failures nobody reports.Tue, 28 Jul 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/schema-tokens-every-sessionhttps://getmcpulse.com/blog/schema-tokens-every-sessionDescriptions and JSON Schema go into the context window at the start of every conversation. Twelve tools can cost 3,000 tokens before anyone asks a question.Tue, 21 Jul 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/dead-toolshttps://getmcpulse.com/blog/dead-toolsEvery server has at least one. It costs schema tokens on every session, competes for the model's attention, and returns nothing, because nobody has ever reached it.Tue, 14 Jul 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/crashed-vs-tool-errorhttps://getmcpulse.com/blog/crashed-vs-tool-errorThe MCP SDK catches whatever your handler throws and turns it into an error result. From outside the handler, a bug and a deliberate error look identical.Tue, 07 Jul 2026 00:00:00 GMTEngineeringhttps://getmcpulse.com/blog/bad-args-is-a-schema-problemhttps://getmcpulse.com/blog/bad-args-is-a-schema-problemWhen arguments fail validation, the instinct is that the model got it wrong. Usually the schema asked for something the model had no way to know.Tue, 30 Jun 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/why-we-hash-argumentshttps://getmcpulse.com/blog/why-we-hash-argumentsTwelve hex characters of a SHA-256 over the arguments, keys sorted. Enough to tell whether two calls were the same, and not enough for anything else.Tue, 23 Jun 2026 00:00:00 GMTPrivacyhttps://getmcpulse.com/blog/not-a-proxyhttps://getmcpulse.com/blog/not-a-proxyA proxy is easier to build and the wrong thing to ship. Directory-listed servers cannot change their URL, and OAuth breaks when traffic is redirected.Tue, 16 Jun 2026 00:00:00 GMTEngineeringhttps://getmcpulse.com/blog/percentiles-cannot-be-summedhttps://getmcpulse.com/blog/percentiles-cannot-be-summedYou cannot average two days of p95 and get the week's p95. Four counters can be added across any range and still tell you the truth.Tue, 09 Jun 2026 00:00:00 GMTEngineeringhttps://getmcpulse.com/blog/sessions-cross-midnighthttps://getmcpulse.com/blog/sessions-cross-midnightBucketed counters are how a dashboard stays fast. But some things are not buckets, and summing them by day quietly double-counts every one that spans two.Tue, 02 Jun 2026 00:00:00 GMTEngineeringhttps://getmcpulse.com/blog/what-a-heavy-response-costshttps://getmcpulse.com/blog/what-a-heavy-response-costsRoughly 2,500 tokens, about a cent, and a chunk of the context window that the rest of the conversation no longer has. The last one is the expensive part.Tue, 26 May 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/writing-tool-descriptionshttps://getmcpulse.com/blog/writing-tool-descriptionsThe model reads your description once, alongside forty others, and decides immediately. Everything that makes documentation good makes a tool description worse.Tue, 19 May 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/the-nightly-passhttps://getmcpulse.com/blog/the-nightly-passThree of the sixteen metrics need to see what came after a call. That is not knowable when the call arrives, so it happens overnight — and the dashboard says so.Tue, 12 May 2026 00:00:00 GMTEngineeringhttps://getmcpulse.com/blog/rollups-exist-so-events-can-be-deletedhttps://getmcpulse.com/blog/rollups-exist-so-events-can-be-deletedEvery call is written twice — once as a row you will throw away, once as a counter you keep forever. The second one is what makes the first one disposable.Tue, 05 May 2026 00:00:00 GMTEngineeringhttps://getmcpulse.com/blog/naming-toolshttps://getmcpulse.com/blog/naming-toolsThe model has your tool name, your description, and nothing else. No source, no docs site, no colleague to ask. The name is doing more work than you think.Tue, 28 Apr 2026 00:00:00 GMTTool designhttps://getmcpulse.com/blog/instrumentation-must-never-break-ithttps://getmcpulse.com/blog/instrumentation-must-never-break-itThe SDK runs inside someone else's production server. Three rules follow from that, and none of them are negotiable — including handing the server back unmodified.Tue, 21 Apr 2026 00:00:00 GMTEngineeringhttps://getmcpulse.com/blog/tool-pairshttps://getmcpulse.com/blog/tool-pairsModels chain tools. When two of yours are always called in sequence, that pattern is usually a design note you have not read yet.Tue, 14 Apr 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/cost-per-sessionhttps://getmcpulse.com/blog/cost-per-sessionNot the monthly bill — that measures how popular you are. Cost per session measures how expensive you are, which is the part you control.Tue, 07 Apr 2026 00:00:00 GMTMeasurementhttps://getmcpulse.com/blog/passing-a-security-reviewhttps://getmcpulse.com/blog/passing-a-security-reviewIf your MCP server touches customer data, someone will ask what your monitoring vendor receives. These are the five questions, and the answers that end them.Tue, 31 Mar 2026 00:00:00 GMTPrivacy