Product
Find recurring requests and the tools people actually reach for before prioritizing the next capability.
MCP server analytics shows who uses your server, which AI clients they use, and whether they come back. Flowlines connects tool calls to users, clients, sessions and use cases, and turns that activity into product and business signals: adoption, buying intent, unmet needs and recurring issues, with notifications for the team that needs to act.
100 analyzed MCP sessions per month, free. No credit card. Connect your assistant, then instrument your server.
Explore the overview, users and use cases of an example customer-operations server.
0sessions
By client · sessions
| User | Client | Sessions | Tool calls | Issues | Last activity |
|---|---|---|---|---|---|
| MRMaya R.Northstar | 14 | 96 | 0 | 2 min ago | |
| PNPriya N.Studio | 38 | 1,284 | 1 | 8 min ago | |
| MDMarc D.Parcel | 21 | 962 | 0 | 1 h ago | |
| AKAiko K.Fieldwork | CCursor | 17 | 731 | 2 | 3 h ago |
| JBJonas B.Orbit | 12 | 506 | 0 | Yesterday |
MCP server analytics measures who uses a Model Context Protocol server, from which AI clients, for which jobs, and whether they come back. Flowlines provides it for teams that ship an MCP server, with users, clients, tools, sessions and returning users in one place.
Infrastructure monitoring tells you whether the service is available. Tool-call telemetry tells you which operation ran and whether it failed. Server analytics adds people, clients, sessions and repeat usage, so you can tell adoption from a busy test script.
Analytics for MCP servers measures how people use an MCP server you ship, and Flowlines is analytics for MCP servers; an analytics MCP server is something else, an MCP server that lets an AI assistant query an analytics product.
If you ship an MCP server and want to know who uses it, which clients call it and what fails, you need analytics for MCP servers. If you want your assistant to answer questions such as last week's signups from your product analytics, you need that product's MCP server. Flowlines does both for its own data: it measures your MCP server, and its own MCP server lets your assistant ask questions about those measurements.
| Question | Analytics for MCP servers | Analytics MCP server |
|---|---|---|
| What it measures | Your MCP server's users, clients, tools and sessions. | Whatever the analytics product behind it already tracks. |
| Who uses it | The team that builds and ships the MCP server. | Anyone who wants an assistant to read their analytics. |
| Example question | How many users came back to my server this week? | How many signups did we get last week? |
Flowlines shows how many distinct people call your MCP server, how many sessions they run and which tools they use, so you can tell real adoption from a test script.
Count users, sessions and tool calls for a fixed window, and exclude internal and test traffic. One script can generate hundreds of calls, while ten people with two sessions each is real adoption. Keep calls without a known user visible as their own group.
| Count | What it tells you | Read it with |
|---|---|---|
| Distinct users | How many people or accounts called the server. | Identity coverage; shared accounts are not individual people. |
| Sessions | How many separate jobs were attempted. | Whether sessions come from your integration or are inferred. |
| Tool calls | How much work the server did. | Users and sessions; retries and pagination inflate it. |
| Internal traffic | Activity from your own team and tests. | Exclude it before reporting adoption. |
Flowlines shows how many users come back to your MCP server in each period, and which AI client they use.
A returning MCP user is someone active in the selected window who was also active in an earlier one; returning users divided by active users gives the returning share. A first call can be curiosity. A second week of activity usually means the server became part of someone's workflow.
| Measure | Definition | Example (synthetic) |
|---|---|---|
| Active users | Users with at least one call in the window. | 24 users in the last 7 days. |
| New users | Active users with no earlier activity. | User 036, 4 sessions. |
| Returning users | Active users who were also active before the window. | User 014, 18 sessions; User 028, 12 sessions. |
| Returning share | Returning users divided by active users. | Report it with identity coverage and the window used. |
Track MCP usage by user and recognize users across clients ↗
Flowlines groups your MCP server's calls by AI client, such as Claude, ChatGPT or Cursor, and keeps calls without a client name in an unknown group.
Client names are reported by the calling application. They tell you where usage comes from, not who the person is. Different clients can attempt different jobs, so compare similar use cases before concluding that one client works better with your tools.
| Client view | Useful for | Caveat |
|---|---|---|
| Calls and users by client | Seeing where adoption comes from. | A client is an application, not an end user. |
| Tools by client | Spotting a tool one client never reaches. | Check discovery and permissions before blaming the tool. |
| Errors by client | Finding client-specific breakage. | Compare the same job; mixes of work differ. |
| Unknown client | Measuring missing metadata. | Keep it visible instead of dropping it. |
Your MCP is a product surface. Measure whether users return, what they need and when they show commercial interest. Flowlines connects these business KPIs to the sessions behind them, so a sales, product or customer team can act on more than a traffic chart.
Use a defined window and the right denominator. Buying-intent incidence is the share of sessions evaluated for buying intent that contain it, not a sales conversion rate. Product-need incidence uses sessions evaluated for product needs. Missing context stays outside the evaluated population, not silently counted as a negative.
| Business question | Measure | Next action |
|---|---|---|
| Are people finding repeat value? | Active users and repeat usage, with identity coverage | Inspect the workflows returning users rely on. |
| Who is evaluating the product? | Sessions with buying intent and buying-intent incidence | Review the account context and notify the commercial team. |
| What should we build next? | Product needs and the sessions they affect | Group unmet requests and inspect the examples before prioritizing. |
| Where does the experience break? | Evaluated outcomes and recurring issues | Find affected users and open the relevant sessions. |
Configure notifications for buying intent, product needs and custom behavior signals. Route findings to Slack, email or a webhook. A user asking about enterprise deployment can matter just as much as a repeated tool failure.
Choose real-time delivery after analysis, or daily and weekly digests. Real-time is not an inline tool-call guard or a fixed latency guarantee: delivery follows analysis, and product-need grouping can add delay. Repeated findings are suppressed until they reopen, worsen or buying intent advances.
KPI analytics and finding alerts are different capabilities. These notifications do not promise arbitrary thresholds on external revenue, billing or CRM data. Configure a behavior your captured sessions can establish, then verify that a synthetic matching session reaches the intended destination.
Try the business-signal examples ↗Compare Flowlines and Datadog notifications ↗Get alerts when MCP tool errors spike ↗Compare alongside PostHog ↗
A rise in MCP traffic can mean new adoption, more work from existing users, or more attempts to finish the same job. The useful next step is to break the increase down by user, tool, client and session.
Suppose a customer-brief workflow moves from three calls to nine. If the same users repeat a search and cannot finish, the extra traffic is friction, not growth. If new teams run more completed briefs, it may be useful adoption. Inspect representative sessions before choosing an explanation.
Find recurring requests and the tools people actually reach for before prioritizing the next capability.
Follow a failing path from the affected session to the calls behind it, rather than starting with a global error count.
Filter users by tool or client to see who returns and who repeatedly encounters issues.
Instrument the MCP server with compatible OpenTelemetry telemetry. Flowlines organizes the observed calls by server, tool and available identity, then connects them into sessions where the context supports it.
The server overview brings together tool usage, clients, users, sessions and use cases. From there, open a tool to see what people use it for, or a session to follow the sequence and how it ended. Period comparisons help you investigate changes; they do not prove a release caused them.
A server does not automatically receive the assistant's complete conversation or final answer. What you can analyze depends on what the integration sends. Inferred sessions stay labeled as inferred.
Follow related calls in MCP session analytics ↗Find failing tools with MCP tool analytics ↗Monitor several MCP servers in one place ↗
The Flowlines plugin gives your assistant the setup workflow. Connecting the assistant does not automatically instrument your MCP server.
MCP OpenTelemetry instrumentation guide ↗MCP observability product overview ↗The MCP analytics measurement guide ↗
Flowlines shows how many distinct people call your MCP server, how many sessions they run and which tools they use. Exclude your own team's traffic: call volume alone can come from one script.
Flowlines shows how many users return in each period. The returning share is users active in the window who were also active earlier, divided by all active users, read with identity coverage.
Flowlines groups calls by AI client, such as Claude, ChatGPT or Cursor, and keeps an unknown group. Client names are reported by the calling application and identify the application, not the person.
Flowlines is analytics for MCP servers: it measures who uses the MCP server you ship. It also has its own MCP server, so your assistant can ask questions about those measurements.
Yes. Flowlines supports adoption analysis, evaluated outcomes, buying-intent incidence and product needs. Read each rate with its evaluated population and inspect the sessions behind it. Purchase intent is not booked revenue, and a tool returning successfully is not proof of a completed business outcome.
Flowlines supports real-time notification delivery after analysis, plus daily and weekly digests. Configure buying-intent, product-need or behavior-signal rules with Slack, email or webhook destinations. Timing depends on analysis and grouping, not a guaranteed per-call delivery deadline.
Yes, if it emits the compatible MCP telemetry Flowlines needs. The Flowlines plugin can guide the instrumentation and verification. Installing the plugin alone does not instrument a server.
Start with the traffic you already have
Connect Flowlines and take the next step with your assistant.