Understand your
agent users

Up to 40% of your API traffic is AI agents. You can't tell which requests are theirs, where they get stuck, or why they churn. General Context reconstructs agent sessions and provides real-time error guidance so agents can self-correct.

Start freeView install
Live Agent Traffic
23% error rate
POST/v1/customersOperator20034ms
POST/v1/paymentsClaude Code2008ms
GET/v1/productsCursor20012ms
POST/v1/paymentsPerplexity400amount must be positive integer
GET/v1/subscriptionsDevin20018ms
POST/v1/checkoutLovable400line_items must be an array
GET/v1/customersOperator429rate limit exceeded
POST/v1/refundsBolt20045ms
POST/v1/checkoutWindsurf400line_items must be an array
GET/v1/invoicesBrowserBase401invalid api key
POST/v1/paymentsClaude Code20022ms
GET/v1/productsManus20014ms
POST/v1/customersOperator20034ms
POST/v1/paymentsClaude Code2008ms
GET/v1/productsCursor20012ms
POST/v1/paymentsPerplexity400amount must be positive integer
GET/v1/subscriptionsDevin20018ms
POST/v1/checkoutLovable400line_items must be an array
GET/v1/customersOperator429rate limit exceeded
POST/v1/refundsBolt20045ms
POST/v1/checkoutWindsurf400line_items must be an array
GET/v1/invoicesBrowserBase401invalid api key
POST/v1/paymentsClaude Code20022ms
GET/v1/productsManus20014ms

Your API has two user bases now

AI agents are your fastest-growing consumers. They hit different endpoints, follow different paths, and fail in ways your dashboards don't show.

Traffic split (last 7 days)
Human 63%Agent 37%

3.2x

more endpoints hit per agent session

4.1x

higher retry rate vs. human users

0

visibility into agent experience

What your logs don't show
1

3 agent frameworks stuck in a retry loop on the same endpoint, all making the same schema mistake

2

Claude completes your payment flow in 3 calls. GPT-4 takes 8 and errors twice. You can't tell them apart.

3

Agent developers silently churn to your competitor's API because your error messages don't explain the fix

Sessions

Reconstruct agent journeys

Raw API calls become coherent sessions. See which endpoints agents hit, in what order, and where confident agents differ from confused ones.

Agent Sessions
AgentWorkflowReqsTimeStatus
Claude Code
POST /customers→POST /payments→POST /checkout→200
30.8sPass
Operator
GET /products→POST /customers→POST /checkout→400→POST /checkout→400
52.4sFail
Devin
GET /products→POST /customers→POST /payments→200
31.1sPass
Lovable
POST /checkout→400→POST /checkout→400
20.4sFail
Agent Comparison
AgentCallsErrorsTimeSuccessBehavior
Claude Code300.8s97%Direct path, reads docs first
Devin401.1s94%Checks catalog before purchase
Cursor411.4s88%Skips validation, retries once
Operator623.1s71%Retry loops on checkout
Lovable320.9s52%Wrong schema, gives up

Compare

See which agents succeed

Compare agent frameworks side by side. See who completes flows efficiently, who retries, and who gives up. Know exactly where to improve your API for the agents that struggle.

Top Failing Endpoints
EndpointError %ErrorsTop Error
POST /v1/checkout68%847line_items must be an array
POST /v1/payments31%204amount must be positive integer
GET /v1/customers24%186rate limit exceeded
GET /v1/invoices12%94invalid api key

Suggested recovery rule

3 agents send line_items as an object instead of an array. Add a recovery rule to guide them to the correct format.

Intelligence

Find where agents get stuck

Surface error patterns across all agent traffic. See which endpoints agents struggle with, which frameworks succeed vs. fail, and what to fix in your API or docs.

Recovery

Help agents self-correct

When agents hit errors, inject guidance into the response so they can fix the request and retry. Define recovery rules once and every agent benefits automatically.

Recovery Flow
REQUEST
POST /v1/checkout
{
  "line_items": { "price_1N": 2 }
}
1,204Guidance sent
847Self-corrected
70%Recovery rate

Built for teams that care about their API

<1ms

Overhead per request

Express middleware. Async telemetry. Your latency budget stays intact.

0

Data leaves your infra

No proxy, no redirect. Your API traffic stays on your servers. We only see metadata.

5 min

To your first insight

npm install, add one line of middleware, deploy. Agent sessions appear immediately.

Two lines. Five minutes to your first insight.

Works with Express, Next.js, or any Node server.

npm install @generalcontext/agentlens
// Track agent sessions + enable recovery guidance.
app.use(agentlens({
  projectKey: "al_srv_...",
  captureErrors: true,
  recovery: { enabled: true }
}));