Skip to main content
Practical GuideDeveloper Tools

API Testing Tools and Agent Readiness: How Postman, Insomnia, and curl Verify Your Score

You do not need a specialized scanner to understand your agent readiness. The tools you already use — Postman, Insomnia, and curl — can test the five things that matter most. Here is the exact workflow, with commands you can run right now.

AH
AgentHermes Research
April 15, 202612 min read

The 5-Step Agent Readiness Testing Workflow

When an AI agent evaluates your API for the first time, it runs through a predictable sequence: check health, test authentication, probe error handling, look for rate limits, and search for a spec. You can simulate this entire sequence with five curl commands. Each step maps directly to a dimension in the Agent Readiness Score.

Step 1

Health check

Verify the API is alive and returning structured responses. An agent's first interaction is always a health or discovery check.

curl -s https://api.example.com/health | jq .

Expected result: 200 OK with JSON body: { "status": "healthy", "version": "2.1.0" }

Step 2

Authentication

Test that auth returns proper 401 with JSON error body (not HTML). Agents need machine-readable rejection to handle auth flows programmatically.

curl -s -o /dev/null -w "%{http_code}" https://api.example.com/protected
curl -s -H "Authorization: Bearer invalid_token" https://api.example.com/protected | jq .

Expected result: 401 with JSON: { "error": "unauthorized", "message": "...", "code": "AUTH_REQUIRED" }

Step 3

Error responses

Send malformed requests and verify errors come back as structured JSON with error codes — not HTML 500 pages or stack traces.

curl -s -X POST https://api.example.com/orders -H "Content-Type: application/json" -d '{"invalid": true}' | jq .

Expected result: 400 with JSON: { "error": "validation_error", "fields": [{ "field": "items", "message": "required" }] }

Step 4

Rate limit headers

Check for X-RateLimit-Limit, X-RateLimit-Remaining, and Retry-After headers. Agents use these to throttle requests and avoid bans.

curl -s -D - https://api.example.com/products 2>&1 | grep -i "rate\|retry\|x-ratelimit"

Expected result: Headers present: X-RateLimit-Limit: 100, X-RateLimit-Remaining: 99, X-RateLimit-Reset: 1714000000

Step 5

OpenAPI spec

Check if a machine-readable API specification exists. This is how agents auto-discover all available endpoints and their schemas.

curl -s https://api.example.com/openapi.json | jq '.info.title, .paths | keys[:5]'

Expected result: Valid OpenAPI 3.x document with paths, schemas, and descriptions for every endpoint.

10 curl Commands That Predict Your Score

Each of these commands tests a specific aspect of your agent readiness. Run them against your own domain and compare the results with what the agent readiness checklist recommends. Together they cover the highest-weighted scoring dimensions.

1. Check for agent-card.json

curl -s https://example.com/.well-known/agent-card.json | jq .

Impact: D9 Agent Experience: +15 points if present and valid

2. Check for llms.txt

curl -s https://example.com/llms.txt | head -20

Impact: D9 Agent Experience: +10 points if present

3. Check for OpenAPI spec

curl -s https://example.com/openapi.json | jq .info

Impact: D2 API Quality: +20 points if valid spec exists

4. Test JSON error handling

curl -s -X DELETE https://example.com/api/nonexistent | jq .

Impact: D8 Reliability: +12 points for structured error responses

5. Check CORS headers

curl -s -H "Origin: https://agent.example.com" -D - https://example.com/api/health 2>&1 | grep -i "access-control"

Impact: D7 Security: +5 points for proper CORS configuration

6. Check TLS certificate

curl -vI https://example.com 2>&1 | grep "SSL\|TLS\|certificate"

Impact: D7 Security: Hard cap at 39/100 without TLS

7. Check response time

curl -s -o /dev/null -w "time_total: %{time_total}s\n" https://example.com/api/health

Impact: D8 Reliability: Sub-500ms is good, sub-200ms is excellent

8. Check Schema.org markup

curl -s https://example.com | grep -o 'application/ld+json' | wc -l

Impact: D6 Data Quality: +8 points for structured data markup

9. Check robots.txt

curl -s https://example.com/robots.txt | head -10

Impact: D1 Discovery: +3 points, critical for crawler access

10. Check status page

curl -s https://status.example.com | head -5 || echo "No status page"

Impact: D8 Reliability: +5 points for public status endpoint

After running these 10 commands: Head to agenthermes.ai/audit and run the full scan. Compare your manual findings with the automated score. The AgentHermes scanner tests 50+ checks across 9 dimensions, including vertical-specific weighting that these curl commands cannot replicate. But you will already know the major gaps before you scan.

Postman vs Insomnia vs curl for Agent Readiness Testing

All three tools can test your agent readiness, but each has different strengths. The best choice depends on whether you want quick spot checks (curl), visual exploration (Insomnia), or comprehensive automated testing (Postman).

curl

Strengths

Fastest for single-endpoint checks. Scriptable. Ships with every OS. Ideal for CI/CD pipelines and quick agent-readiness spot checks.

Agent Readiness Relevance

Tests exactly what an agent sees: raw HTTP request and response. No UI abstractions. The closest simulation of an agent's actual experience.

Postman

Strengths

Import OpenAPI specs to auto-generate test collections. Visual response inspection. Environment variables for staging vs production. Team collaboration on test suites.

Agent Readiness Relevance

Collection Runner can test every endpoint in sequence — mimicking an agent's discovery flow. Pre-request scripts simulate auth token refresh cycles agents perform.

Insomnia

Strengths

Lightweight alternative to Postman. Native OpenAPI support. Environment switching between test and production. Plugin ecosystem for custom auth flows.

Agent Readiness Relevance

Environment variables let you test the same endpoints against dev, staging, and prod — verifying consistency across environments, which agents expect.

The Postman Agent Readiness Collection

If you have an OpenAPI spec, Postman is the most powerful tool for comprehensive agent readiness testing. Here is the workflow we recommend.

1

Import your OpenAPI spec

File > Import > paste your openapi.json URL. Postman auto-generates a request for every endpoint with example parameters pre-filled from the schema.

2

Create environment variables

Set up two environments: "Production" and "Staging." Variables: base_url, api_key, auth_token. This lets you verify that both environments return consistent responses — which agents expect.

3

Add test scripts to every request

Postman tests verify: response is JSON (pm.response.to.be.json), status code is correct, response time is under 500ms (pm.response.responseTime < 500), and required fields exist in the body.

4

Run the Collection Runner

Execute all requests in sequence. This simulates an agent discovering your API: first the spec, then auth, then iterating through every endpoint. The runner report shows which requests fail.

5

Check the results against the checklist

Compare your Collection Runner results with the agent readiness checklist. Every failing request is a dimension score reduction. Fix the failures, re-run, and verify improvement before scanning with AgentHermes.

The key insight is that Postman Collection Runner mimics agent behavior. An AI agent discovering your API goes through the same sequence: fetch spec, authenticate, iterate endpoints, check responses. If your collection passes cleanly in Postman, your API will score well when agents evaluate it. If requests fail, those are exactly the failures agents will encounter.

For businesses that already have an OpenAPI specification, this workflow takes under 30 minutes and gives you a reliable preview of your Agent Readiness Score before you run the official scan.

Frequently Asked Questions

Can I really predict my Agent Readiness Score with curl?

You can approximate 60-70% of your score with the 10 curl commands in this guide. They test the most heavily weighted dimensions: API quality (D2), security (D7), reliability (D8), and agent experience (D9). The remaining 30-40% comes from deeper analysis that AgentHermes performs — like evaluating response schema consistency, documentation quality, and vertical-specific scoring weights. Run the curl tests first, then run the full AgentHermes scan to see the complete picture.

What is the most impactful single test I can run?

Check for an OpenAPI spec: curl -s https://yoursite.com/openapi.json. If this returns a valid OpenAPI 3.x document, your D2 API Quality score jumps significantly. If it returns 404, your API is undiscoverable to agents — they cannot learn what endpoints exist, what parameters they accept, or what responses to expect. The OpenAPI spec is the single highest-leverage artifact for agent readiness.

Should I use Postman or Insomnia?

For agent readiness testing specifically, Postman has an edge because of its Collection Runner and ability to import OpenAPI specs directly. You can import your spec, auto-generate requests for every endpoint, and run the full collection — simulating what an agent does during discovery. Insomnia is lighter and faster for quick manual testing. Both work well. If you have neither, curl covers the critical tests.

How do I test if my error responses are agent-friendly?

Send requests that should fail: missing auth, invalid JSON body, nonexistent resource IDs, wrong HTTP method on an endpoint. For each, check that the response is JSON (not HTML), includes a machine-readable error code (not just a human message), and returns the correct HTTP status code. If any error returns an HTML page or a 200 status with an error message in the body, that is an agent readiness problem. Agents cannot reliably parse HTML error pages.

What does the AgentHermes scan test that curl cannot?

AgentHermes tests 9 dimensions across 50+ checks including: vertical-specific scoring weights (a restaurant is scored differently than a SaaS product), Schema.org structured data parsing, MCP server detection and protocol validation, agent-card.json schema compliance, cross-referencing multiple endpoints for API consistency, documentation quality scoring, and comparison against 500+ scanned businesses. The curl tests cover the binary checks — does this exist or not. AgentHermes evaluates quality and completeness.


Done testing manually? Run the full scan.

The AgentHermes scanner tests 50+ checks across 9 dimensions with vertical-specific scoring. Compare your curl results with the automated score in 60 seconds.


Share this article: