FAQ

Storage above your plan's included quota is metered at $2 per GB per month. We email you at 80% and 100% of your included quota so there are no surprises.

Yes. We use Dodo Payments as Merchant of Record, which collects and remits VAT, GST, and sales tax across 150+ jurisdictions on your behalf. You don't need to register or remit.

Yes. Cancel from the billing page; you keep access through the end of the current billing period, then drop to the free tier (or your endpoints are archived if you exceed free-tier limits).

Enterprise plans cover 500 GB+ storage, BAA, dedicated support, custom integrations, and SLA. Email matt@mrdula.solutions to talk.

Native parsers for CSV, TSV, JSON, NDJSON, SQL dump, Excel, Markdown, TXT, HTML. Workers AI document conversion for PDF, DOCX, PPTX, RTF, EPUB. Image OCR for PNG, JPG, JPEG, WEBP, GIF, BMP, TIFF. Whisper turbo for MP3, WAV, M4A, OGG, FLAC, OPUS, WEBM. Plus URL extraction (markdown / structured JSON / crawl) for any web source.

Free: 25 MB per file. Pro: 100 MB. Scale: 500 MB. Audio files have a tighter cap of 25 MB regardless of plan because that's where Whisper turbo's request body limit sits — chunking for longer audio is on the post-launch roadmap.

On Pro+ endpoints with the "Allow agent writes" toggle on, schema-gen produces insert_<table>, update_<table>_by_id, and delete_<table>_by_id tools alongside the read tools, plus four built-in ingest tools (ingest_url, ingest_file, request_upload_url, ingest_uploaded_file). Agents need a separate mcp:write OAuth scope to call them — your consent screen has a checkbox to grant it. Writes are async via a job queue and every one is audit-logged.

Yes. We implement the MCP-spec OAuth 2.0 + Dynamic Client Registration flow (RFC 8414, RFC 7591, RFC 7636 PKCE). Paste your endpoint URL into Claude.ai's connector setup, click Allow on the consent screen, done. Same for Cursor and any modern MCP client. Claude Desktop's static-token config still works too.

Anything that speaks MCP. Tested first-party: Claude.ai web, Claude Desktop, Claude Code (CLI), Cursor, MCP Inspector. Compatible by spec: ChatGPT (when MCP shipped), Grok, Perplexity, OpenClaw, Manus, the Claude Agent SDK, LangChain via MCP, and any custom HTTP/JSON-RPC client.

Yes. Open /endpoints/<id>, scroll to the Danger Zone, type the slug to confirm. We hard-delete the D1 row immediately (frees the plan-cap slot), and the worker tears down the per-tenant Neon project + R2 prefix asynchronously within a few seconds.

Yes. Settings → Delete account. We hard-delete every endpoint, every R2 file, every audit log row, every token, every OAuth client we issued, within 30 days. Payment ledger rows are retained for 7 years for tax compliance — that's the only exception.

Endpoint content (Postgres rows, R2 archives): until you delete the endpoint or account. Account data: until you delete. Write/ingest job rows: 30 days. Audit log: 12 months. Operational logs: 30 days. Magic-link tokens: 15-min TTL + 24-hour cleanup. Payment ledger: 7 years.

No. Customer content (uploaded files, ingested URLs, Postgres rows, R2 archives) is never used to train models. Workers AI processes content during ingest (PDF→markdown, audio→transcript) but Cloudflare's Workers AI does not retain content for training.

You keep access through the end of the current billing period. After that, your account drops to the free tier; if your usage exceeds free-tier limits (1 endpoint, 500 MB storage, 1,000 MCP calls/month) those endpoints are archived (set to status='archived', tools/list returns empty, no new ingest accepted) until you upgrade again or delete them. Nothing is hard-deleted until you explicitly delete the account.