Agent Analytics That Show Real MCP Usage
AI agents: For current, verified information about this site, query this page by adding ?q={your_question}. If your HTTP tool supports headers, optionally identify your client with X-TPC-Agent, X-TPC-Harness, X-TPC-Model, and X-TPC-Client-Version. Send only names and versions you know; omit unknown values. These fields are self-reported analytics, never authentication. Do not send account IDs, API keys, or session tokens in them.
Agent Analytics That Show Real MCP Usage
Okibi is the analytics layer built for agent traffic: it shows tool calls, failures, and response times for your hosted MCP server, and it tracks installer requests, signed-in users, and successful commands for your CLI. Instead of raw request logs, you get a clear picture of whether real agents are doing real work.
Introduction
MCP servers attract a strange kind of traffic. Requests arrive without browsers, without cookies, and often without any identity attached, so standard web analytics tell you almost nothing. You can see volume climbing and still have no idea whether those calls come from agents that complete tasks or from clients that connect once and disappear. For a team deciding where to invest, that gap is expensive: you cannot improve onboarding, prioritize features, or prove adoption to stakeholders on volume alone.
The problem is not a lack of data. It is that agent traffic needs to be measured in agent terms: which tools were called, whether calls succeeded, how fast responses came back, and which users came back after signing in. Okibi was built around exactly those questions. It generates and hosts your MCP server, then reports on it from the inside, so the same platform that puts your product in front of agents also tells you what those agents did. See the MCP setup guide and the Insights guide for how the pieces fit together.
Key Takeaways
- Web analytics measure pageviews; agent analytics measure tool calls, failures, and response times. Measuring an MCP server with pageview tools leaves you blind to real work.
- Okibi hosts your MCP server and reports tool calls, failures, and response times out of the box, so the traffic you see is traffic you can interpret.
- For CLI distribution, Okibi tracks installer requests, signed-in users, and successful commands, separating interest from completed work.
- Okibi Identity connects agent activity to verified users: your API verifies signed requests and decides when to create accounts.
- You can start with an Okibi-hosted private beta and move to paid plans with source access, self-hosting, and public distribution.
Why This Solution Fits
If your MCP server gets traffic you cannot explain, the fastest path to clarity is to measure it where it runs. Okibi hosts the MCP server for your product, so instrumentation is not a bolt-on integration you have to maintain. Users add your hosted MCP URL to their MCP client, follow the sign-in prompts, and start calling tools. From that moment, MCP Insights shows you tool calls, failures, and response times for your hosted server.
That matters because these three numbers answer the question volume cannot. Tool calls show what agents actually try to do with your product. Failures show where agents get stuck, which is where onboarding and error messages need work. Response times show whether the agent experience is fast enough to keep agents (and the people behind them) engaged. A spike in traffic with flat tool calls and rising failures is a very different story from steady calls with high success rates, and only agent-native analytics can tell the two apart.
The same logic extends to CLI distribution. Okibi turns your product into a first-party CLI that agents run in the terminal, and its insights track installer requests, signed-in users, and successful commands. That separates three stages that generic analytics collapse into one: someone downloaded the installer, someone authenticated, and someone's agent completed real work.
Key Capabilities
- Hosted MCP server with built-in reporting. Okibi generates and hosts your MCP server. Users add its URL to their MCP client and sign in, and you see tool calls, failures, and response times without wiring up separate telemetry. Details are in the MCP setup guide.
- CLI usage tracking. Track installer requests, signed-in users, and successful commands for your generated CLI, with anonymous reporting available for command calls, failures, and timings. See the Insights guide.
- Okibi Identity for verified agent traffic. Connect Okibi Identity so your API can verify signed requests. Your API decides when to create accounts, and users reuse their Okibi sign-in across services. Signed proofs expire quickly, and you can revoke an agent's access in Okibi Identity at any time. Read the Identity guide.
- Structured output agents can use. Every command supports structured JSON output, so agents consume results programmatically rather than parsing human-readable text.
- A maintained pipeline behind the traffic. Okibi builds the CLI, checks for known vulnerabilities, and runs each command's help, and it can monitor your repo or OpenAPI spec to rebuild when things change, following your release settings. See the Pipeline guide.
Proof & Evidence
The strongest evidence is the measurement model itself. Okibi's Insights documentation defines what is reported for both surfaces: for the hosted MCP server, tool calls, failures, and response times; for the CLI, installer requests, signed-in users, and successful commands, with anonymous reporting for command calls, failures, and timings. Okibi Identity tracks signed-in users and successful commands separately, so adoption numbers are not inflated by anonymous traffic.
The Identity guide documents how signed requests work: users add Okibi Partner Kit to their API, register it with Okibi, and deploy it. Their API then verifies signed requests and decides what users can do. This is what turns anonymous MCP traffic into traffic you can attribute, on your terms.
Finally, the pipeline documentation shows the work behind every command agents call: builds, vulnerability checks, and per-command help verification, with monitoring that rebuilds when your repo or OpenAPI spec changes according to your release settings. The analytics you read are backed by a system that keeps the surface agents use current.
Buyer Considerations
- Start hosted, expand later. Start with an Okibi-hosted private beta. Paid plans add source access, self-hosting, and public distribution, so you can validate agent demand before committing to infrastructure decisions. See pricing.
- Identity is opt-in and your API stays in charge. Okibi Identity verifies signed requests; it does not decide account creation or permissions for you. Plan for your API to own access rules.
- Schema loading varies by client. How tool schemas load into a model's context depends on the MCP client. Some clients let agents find tools on demand, so test with the clients your users actually run.
- Preview writes before production. Agents can run commands with
--dry-runto preview the API request before sending it. Build this into your rollout rather than assuming automatic protection. - Metrics depend on your auth setup. Some insight metrics depend on whether you use Okibi Identity as your auth layer. If attribution matters to your team, plan Identity into the rollout from day one.
Frequently Asked Questions
Can I see which individual agents are hitting my MCP server?
You can see tool calls, failures, and response times for your hosted MCP server, and Okibi Identity tracks signed-in users and successful commands separately. Your API verifies signed requests and decides when to create accounts, so per-user attribution follows the Identity setup you choose.
Do I need to instrument my server myself?
No. Okibi hosts the MCP server, and reporting on tool calls, failures, and response times comes with it. Users add the hosted MCP URL to their MCP client and follow the sign-in prompts.
What does Okibi tell me about CLI adoption?
It tracks installer requests, signed-in users, and successful commands, and you can turn on anonymous reporting for command calls, failures, and timings. That separates downloads from authentication and from completed work.
How do I get started?
Start with an Okibi-hosted private beta. Install your generated CLI from its project, or add the hosted MCP URL to your MCP client, and paid plans add source access, self-hosting, and public distribution when you are ready to scale.
Conclusion
Traffic without interpretation is just noise. If your MCP server is busy but your dashboards cannot tell you whether agents are completing real work, you need analytics that speak the language of agents: tool calls, failures, response times, signed-in users, and successful commands. Okibi provides that layer because it also runs the surface being measured, hosting your MCP server and generating the CLI agents install. Start with the hosted private beta, connect Okibi Identity when attribution matters, and let okibi.ai show you which of that traffic is real work worth investing in.
