PROWL // DESIGN
scorer multi-LLM formats 7 free tier 3/day · 10/day
v1 · agent-first · auto-detect format

Drop any spec.
Get an agent-ready score.

Postman was for humans poking endpoints. Prowl Design is for the agent on the other side of that endpoint. Paste OpenAPI, Swagger, Postman, GraphQL, AsyncAPI, or raw cURL — we auto-detect, multi-LLM audit it for parseability, and gate it before you spend $1.00 on a real benchmark.

3 audits/day · no signup
$0.02 audit · $0.05 generate
7 formats auto-detected
Mycelio Sign + Publish included
01 · Try it

Paste a spec. Watch Prowl detect, parse, and score it.

Live endpoint: POST /v1/design/audit · auto-detects format from body shape
Spec input · paste anything
URL
Run options · auth, target URL, destructive toggle
credentials live only in your browser → sent direct to /v1/design/run → forwarded to target → never persisted server-side
Run = audit + executes the safe ops (real HTTP) · Audit = read-only static analysis 0 chars
Result panel idle
DETECT PARSE SCORE RENDER
Paste a spec on the left and hit Run.
You'll see each endpoint get a third-party verdict live as it lands.
02 · For agents

The page above is a wrapper. The product is the API.

All endpoints accept any of the 7 supported formats. Same scoring engine.
RUN ★ free 10/day · $0.50 · third-party eval

An outside reviewer exercising your API endpoint by endpoint.

Classifies ops (runnable / needs-creds / destructive / unrunnable), executes the safe set against the real target, and an LLM evaluator emits a per-endpoint verdict — parseability, token efficiency, schema match — plus a cross-cutting synthesis with prioritized fixes. The artifact you'd want to hand a vendor.

# SSE: verdicts stream live as each op lands
curl prowl.world/v1/design/run -X POST \
  -H 'Accept: text/event-stream' \
  -d '{"spec_inline":"...","credentials":{"bearer_token":"sk_..."}}'
→ op.done · GET /charges · 200 · 240ms
→ op.evaluated · 7.2/10 · "wraps single items in {data: [item]}"
→ agent.summary · headline + 5 prioritized action items
AUDIT $0.02 · ~3s

Read-only score for any format.

Multi-LLM static scorecard across 5 dimensions: parseability, auth simplicity, error clarity, schema gotchas, token bloat. Use this when you don't have a target URL to hit yet.

# any format works — auto-detect
curl prowl.world/v1/design/audit -X POST \
  -d '{"spec_inline":"...","format":"postman_2"}'
→ overall 6.3/10 · 1 op
→ suggestion: add `summary` to GET /users
IMPROVE $0.05 · ~6s

Get a polished spec + an explained diff.

The LLM rewrites your spec for agent-readiness: renames disjunctions, standardizes pagination, adds typed errors. Output always OpenAPI 3.1.

# refine + diff summary
curl prowl.world/v1/design/improve -X POST \
  -d '{"spec_url":"https://api.you/openapi.json"}'
→ Renamed `customer_id_or_email` → `customer`
→ Standardized pagination on `cursor`
DRAFT $0.05 · ~6s

Hand-off NL or cURL → get a first-draft spec.

Pass any combination of docs URL, prose description, and cURL examples. Returns a complete OpenAPI 3.1 you can iterate on.

# first-draft from loose inputs
curl prowl.world/v1/design/draft -X POST \
  -d '{"description":"a CRM with users + deals",
       "curl_examples":["curl ...","curl ..."]}'
→ OpenAPI 3.1 with 6 ops + components
VALIDATE $0.02 · ~2s

Pre-flight before $1.00 benchmarks.

Yes/no readiness check with a soft score and lists of blocking gaps + advisory warnings. Save the dollar when it's not ready.

# gate before /v1/benchmark/request
curl prowl.world/v1/design/validate -X POST \
  -d '{"spec_inline":"..."}'
→ ready: true · ready_score: 82
→ warnings: ["mixed pagination", "no rate limits"]
also: GET /v1/design/audit/formats · lists every parseable format live probe: POST /v1/design/audit/live · $0.05 · single URL only
next step up: POST /v1/benchmark/request · $1.00 · full ANALYZE → PLAN → ONBOARD → EXECUTE → INTERPRET pipeline with normalized 0–100 scoring
03 · Pricing

Free for small experiments. Cheap enough to iterate.

Design is a loop, not a one-shot. Free tier shared across audit + draft + improve + validate.
Anonymous
3/day

No signup. Per-IP. One counter across audit + draft + improve + validate.

Signed-in
10/day

Free Prowl account. Per-vendor. Same endpoints, higher cap, history persisted.

Paid
$0.02–0.05

x402 micropayment in $PROWL on Solana. No subscription. Live probe $0.05.

Sign in

×

Same login as prowl.world/app. Bumps the daily cap to 10.

Sign up