Pine Check API v0.2.1

Every time your AI writes or edits Pine code, one call tells you where the backtest will lie — future data, unconfirmed higher-timeframe values, fills and costs live trading will not reproduce — with the line, the rule, and what was not checked.

Try it (no key, 30 checks a day)

curl -X POST https://pine.pokt-agent.com/v1/integrity -H "content-type: application/json" \
  -d '{"source":"//@version=6\nindicator(\"t\")\nh = request.security(syminfo.tickerid, \"D\", high, lookahead=barmerge.lookahead_on)\nplot(h)"}'

→ FAIL, P201 on line 3: the daily high is read from the future on historical bars. Fix: high[1] inside the call.

Endpoint	Purpose
`POST /v1/integrity`	backtest-vs-live checks only (P201–P212)
`POST /v1/lint`	adds syntax and version checks (v4 remnants in v5/v6 scripts, brackets, strings, limits)
GET /v1/rules	rule catalogue
GET /openapi.json	OpenAPI 3 spec

Plans

Plan	Price	Included
Free trial	0	30 checks / day, no key
Pilot	₩19,000 (≈ $14)	100 checks / 14 days
Solo	₩49,000 (≈ $35) / month	300 checks / month
Team	₩99,000 (≈ $71) / month	1,000 checks / month

Send the key as Authorization: Bearer pk_.... Responses carry X-Quota-Remaining and X-Request-Id. Get a key: open a request with the plan you want; we reply there with payment details. Pilots are invoiced while card checkout is being set up.

What it is, and is not

Pine v5 and v6: all rules. v3/v4: a version warning only.
Static checks: it does not compile or run the script. CLEAN means none of the listed patterns were found — not that a strategy is safe, profitable or repaint-free. Not trading advice.
Measured on 696 open-source scripts: 0.8% of v5/v6 scripts fail, and each failure checked by hand was a real issue.
Billing: a call is billed only when the check runs and returns a result. Errors and malformed input are free.
Your code is not stored: the log keeps a hash, the size and the result counts. Never used for training, never sent to an LLM.

Service id pine-script-lint-v1, also served on Pocket Network.