Monthly plans for the
MCP compression gateway.

gotcontext chunks the document, embeds each chunk with a local ONNX sentence-transformer (no external embedding API call), builds a similarity graph, scores nodes with PageRank, and returns the top-K chunks in document order as a compressed skeleton. Token-reduction depends heavily on document size, fidelity setting, and embedding tier. See /v1/global-savings for the live rolling average across production traffic and use the Savings section below to project against your own usage.

When you reach your plan’s monthly compression limit, /v1/compress returns HTTP 429 with a Retry-After header set to the start of your next billing cycle. Upgrade in the dashboard and the new ceiling applies immediately. We email at 80% and 100% of quota. Overage pricing is on the Phase-2 roadmap; today it’s hard-stop-then-upgrade.

No. gotcontext is a compression layer. We don’t fine-tune or train any models on customer inputs. The compression engine is deterministic ONNX inference plus PageRank; there is no learned component that sees your text. Inputs to /v1/compress are processed in-memory; payloads are not retained beyond the request-response lifecycle (Business/Enterprise: zero-retention mode in self-hosted Docker keeps prompts entirely inside your VPC).

None during compression. The pipeline runs locally with ONNX/SBERT/TF-IDF embeddings. No OpenAI, Anthropic, or Google API call is made by gotcontext. Your downstream LLM (the model you send the compressed output to) is entirely your choice and your contract. We do estimate cost savings against the model you name in the request body, using each provider’s published per-token rates.

Yes, as a capability of the compression API. POST /v1/compress accepts an optional cache_policy field that leaves a static prefix (for example, your system prompt) byte-for-byte untouched and compresses only the dynamic remainder, then returns cache_breakpoints for Anthropic, OpenAI, or Gemini so your downstream provider can hit its own native prompt cache on repeat calls. Both SDKs ship an apply_anthropic_breakpoints helper that stamps the returned breakpoint onto an Anthropic messages array for you.

Yes. Plan switches between Free, Pro, Team, and Business take effect immediately on upgrade and at the start of the next billing cycle on downgrade. Annual contracts on Business+ run for the full term; mid-term changes are prorated against the upgrade differential.

Dashboard → Billing → Cancel subscription. Monthly plans end at the next renewal date; you keep access for the remainder of the paid period. Annual plans run to term and do not auto-renew when canceled. On termination all customer data (API keys, project state, audit logs) is destroyed within 30 days; certificate of destruction available on request for Business+.

Production runs in AWS us-east (via Fly.io iad region). EU residency is available on Enterprise by request. Self-hosted Docker (Business+) runs entirely inside your VPC. gotcontext sees only metered usage events, never your prompts.

Available on Team and above, the async batch queue lets you submit up to 50 documents in a single /v1/compress/batch request. Jobs are processed in the background; you receive webhook notifications and can stream progress via GET /v1/batch-queue/stream (SSE) from the dashboard’s Queue Monitor.

Free gets the 23 core compression tools (ingest_context, read_skeleton, search_semantic, gc_pre_flight, etc). Paid plans get the full MCP tool catalog covering compression, semantic memory, code analysis, multimodal, and ACE workflows. The complete tools/list manifest is enumerable from any authenticated MCP session at https://api.gotcontext.ai/mcp.

Annual billing on Pro ($39/mo) and Team ($79/mo) is a flat 20% off the monthly rate, framed as "get 2.4 months free." Business is annual-invoice only. Enterprise is custom contract. The annual term is non-cancellable mid-cycle, but unused months are refunded prorated if you cancel within the first 30 days.

Free, Pro, and Team are best-effort with no contractual SLA, which matches industry norm for self-serve tiers (Vercel Pro, Cloudflare Workers, Supabase Pro: no SLA either). Business carries a 99.5% monthly uptime SLA with a 10% / 25% / 50% service-credit schedule (claim-based, capped at one month of fees). Enterprise uptime and credit terms are negotiated per contract. A public status page runs at gotcontext.ai/status; per-component monitoring and 90-day uptime history are on its roadmap. Required reading before signing for any tier with an SLA.

On termination all customer data (API keys, project state, audit logs, cached embeddings) is destroyed within 30 days. Certificate of destruction available on request for Business and Enterprise customers. Self-hosted Docker (Business+) keeps payloads inside your VPC the whole time. gotcontext sees only metered usage counts, never your prompts; deletion is your own infra operation.