Intelligence Pricing.

Free gives you 1,000 requests a month at 60 requests a minute, on the same dataset every paid tier reads: the AI Role Taxonomy across 3,445 roles and 81 cities, the calculate and search endpoints, and the published BLS wage benchmark that travels beside every estimate. It needs an account and an API key, because the monthly limit is per user and an anonymous request cannot be counted against a user. It does not need a card, and there is no clock on it.

You do not need one, because the free tier is permanent. It gives you 1,000 requests a month at 60 requests per minute on the same dataset every paid tier reads, with an API key but no card and no clock on it. That is the evaluation path, so there is no 14-day trial to start or forget to cancel. When you need more volume, Pro and Ultra bill from the day you subscribe, and canceling drops you back to free rather than cutting you off.

Pro ($99/mo) adds throughput (300 requests a minute, 250,000 a month), the MCP server, the /lineage provenance endpoint, restricted API keys with per-engine scopes, and priority support. Ultra ($199/mo) adds 1,500 requests a minute and 500,000 a month, the company leveling catalog, the annual compensation reports, dedicated support, and custom data requests. Every response on every tier carries its methodology version.

The MCP server at /api/v1/intelligence/mcp is a JSON-RPC 2.0 endpoint exposing six locked tools: list_capabilities, analyze_compensation, analyze_skills, analyze_market, discover_roles_and_cities, find_adjacent_opportunities. Every tool returns a Decision-Ready Response with a quotable answer, primary + supporting data, methodology citation, follow-up suggestions, and honest limitations. Available on Pro and above using your same API key. Drop into Claude Code, ChatGPT, or any MCP client.

Yes: on Pro and above. Create a key with only the scopes it needs: intelligence:read, skills:read, market:read, compensation:read, company_data:read, intelligence:lineage. A key granted only skills:read cannot reach the company catalog. This is the recommended pattern for production keys: limit blast radius if a key is ever leaked.

Every salary data point carries a stable data_point_id (e.g. ai-engineer:san-francisco:median_base:2026-Q2). The /lineage endpoint resolves that ID into the full provenance trail: every source that contributed, with fetch timestamps, contribution weights, and the reconciliation rule applied. Pro tier with the intelligence:lineage scope.

Every API call to any endpoint counts as one request, against both the per-minute rate limit and the monthly quota. Cached responses (same query within 60 seconds) do not count against your limit. Idempotent POST replays (same Idempotency-Key + same body) return the cached response with X-Idempotent-Replayed: true and don't count.

You will receive a 429 response with X-RateLimit-Remaining and X-RateLimit-Reset headers and a retry_strategy field in the error envelope. Exceeding the monthly quota returns a 429 too, naming the quota and the reset date. Upgrade for more throughput, or add backoff using the retry_hint. @orbyt/cli auto-throttles when remaining approaches zero.

The role taxonomy updates quarterly, because role identities are slow-moving. The published BLS wage benchmarks update when BLS publishes a new OES release, and every response names the release it is quoting. Every response also carries methodology_version, so you always know exactly what you are citing.

Yes. All paid plans include commercial use rights. Cite 'Orbyt Intelligence' in published work. Contact us for redistribution licensing.

Yes. Paying yearly saves 15% or more against the monthly price on both paid tiers.

Yes. Your subscription ends at the end of the current billing period (no termination fee, no minimum commitment). Your API keys keep working: they drop to the free tier's limits when the period ends, and go back to your paid limits the moment you resubscribe.

API key authentication, on every tier including free. Generate a key in your dashboard, optionally restrict it to specific scopes. No OAuth flows, no token refresh. One Authorization: Bearer header, every request.