How to get your Parallel API usage, balance and spend programmatically
What the Parallel API exposes
| Signal | Parallel |
|---|---|
| Balance | No |
| Spend history | No |
| Cost breakdown | No |
| Plan and quota | No |
| Cost on each response | No |
| Rate-limit headers | None |
| Credential | API key |
No billing API, so measure from your side
Argus prices each Parallel call from what the SDK records: the endpoint and mode, priced at Parallel's per-request rates (Search $1–$5 per 1,000, the Task API by processor tier).
Credential. A Parallel API key, sent as x-api-key. Argus validates it with a turbo search, the cheapest mode.
Gotchas
- Only POST requests that create work count against the rate limit; polling a task's status or result doesn't.
- Task API cost varies by processor tier by orders of magnitude. Record which processor each call used.
How Parallel charges
Usage-priced per request: Search $1–$5/1k, Task API by processor tier ($5–$2,400/1k tasks). Prepaid/dashboard billing; no public balance/usage API.
At zero. Prepaid/dashboard billing (platform.parallel.ai); no public balance/usage API. Qualified startups may receive up to $250 free credits.
- 600 requests per minute, per key. Only POST (creating) requests count; status/result polls don't.
- 2,000 requests per minute, per key.
Or let Argus track it
Paste the key once. Argus keeps the ledger, and your coding agent reads it through the MCP server.
- With the SDK, itemizes each call by endpoint and prices it at Parallel's per-operation rates.
- Emails you on low balance, days to zero, a spend spike, or projected overage.
- Hands all of it to your coding agent over MCP, so it can ask what a run cost or how much runway is left without leaving the editor.
> what has parallel cost today, and how long until it runs out?
get_overview → parallel: mtd spend, burn per day, days to zero
get_request_costs { "provider": "parallel", "limit": 25 }
get_advisories → credits_low, spend_spike, rate_limit_low …Questions
- Does Parallel have an API for its balance?
- Parallel has no public balance or usage endpoint. Billing is prepaid in the dashboard at platform.parallel.ai, and responses don't report cost.
- Can I get Parallel spend per day from the API?
- No. Count calls and price them yourself, or let the Argus SDK do it.
- What happens when Parallel credits run out?
- Prepaid/dashboard billing (platform.parallel.ai); no public balance/usage API. Qualified startups may receive up to $250 free credits.
- What key does Argus need for Parallel?
- A Parallel API key, sent as x-api-key. Argus validates it with a turbo search, the cheapest mode.
Related
Pricing and limits verified 2026-07-21 against docs.parallel.ai/getting-started/pricing, docs.parallel.ai/getting-started/rate-limits, docs.parallel.ai/search/search-quickstart. Endpoints are the ones the Argus Parallel connector calls.
Connect a provider key, point Claude Code or Cursor at the Argus MCP server, and ask what the last run cost.