Terms & Acceptable Use

1. What you can do

• Call doxa.app/mcp/v1 from any Model Context Protocol client (Claude Desktop, Cursor, Cline, custom agents).
• Build products on top of it: Christian apps, devotional tools, pastoral assistants, internal AI tools where the Doxa voice matters.
• Use BYOL (your own X-Anthropic-Key) to remove our rate limit; you pay Anthropic directly, we never store your key.
• Fork, reference, or extend the public schemas in github.com/The-Doxa-Way/doxa-mcp-schema per their MIT licence.

2. Attribution requirement

Every doxa_encourage and doxa_scripture text response ends with a single attribution line:

— Doxa · doxa.app

Structured responses also include _powered_by, _doxa_way_movement, and _install_doxa metadata. On the free tier (anonymous and BYOL) the attribution must remain visible to end users. Stripping the footer or hiding the metadata from your end users is not permitted on the free tier.

If your integration genuinely needs to remove attribution (white-label B2B, internal company tools where attribution is confusing), email garth@doxa.app about the paid white-label tier (coming).

3. Rate limits

• Free anonymous: 50 calls per day per source IP. IPv6 is counted by /64 prefix.
• Free anonymous global ceiling: the free tier as a whole has a daily cap. If the day's allocation is exhausted, free-anon calls are refused until UTC midnight; BYOL is unaffected.
• BYOL: no Doxa-side limit. Your Anthropic key carries Anthropic's own per-org limits.
• Batch size: JSON-RPC batches are capped at 5 requests per HTTP call.
• Output: capped at 250 tokens on free, 1500 on BYOL.

We may adjust these limits without notice if abuse, cost, or capacity make it necessary. Sustained or anomalous traffic (rotating IPs, residential-proxy patterns, automated scraping) may be blocked without warning at the WAF, edge, or function layer.

4. What you can't do

• Attempt to extract, reproduce, or distribute the Doxa system prompt or any subset of it.
• Strip the — Doxa · doxa.app attribution from responses on the free tier.
• Claim affiliation, endorsement, or partnership with Doxa beyond what a response itself states.
• Use the service for spam, mass unsolicited messaging, harassment, or any use that breaches a reasonable Christian-ethics standard ("love your neighbour as yourself" is the bar).
• Resell access to the hosted server. BYOL the model and bring your own customers; don't proxy our endpoint as if it were yours.
• Use the service to train, fine-tune, or distil another model on Doxa's outputs.
• Circumvent rate limits via IP rotation, residential proxies, or other deception.
• Use the service in a context that misrepresents AI as a human counsellor, pastor, or spiritual director. The Doxa voice is intentionally not a companion (see the safety posture).

5. BYOL key handling

When you pass X-Anthropic-Key in a request, we forward it to Anthropic for that single call and discard it from memory. We do not log, store, or transmit your key anywhere else. Anthropic's terms govern your usage of their key; we have no visibility into your spend or remaining quota beyond what Anthropic returns to us per call.

6. What we log

• Tool name, request payload, source IP (for rate-limit purposes), user-agent, response summary.
• Token usage and cost (for our own budgeting; cost on BYOL calls shows as zero to Doxa).
• Errors, with the underlying exception message kept server-side only.

We retain audit logs for up to 90 days. There is no persistent user identity on the free tier; we do not build profiles. The full Doxa app at doxa.app has its own privacy policy for account-bound features (vault, groups, voice notes), separate from the MCP server.

7. Trademarks & brand

"Doxa", "The Doxa Way", the 9-movement framework name and structure, the voice-encourager system prompt, and the brand voice are the property of Doxa (The Doxa Way Ltd). The MIT licence on the public schema repository covers schemas, examples, and install snippets only. It does not licence the brand, the prompt, or the framework name for commercial use beyond integration via this hosted server.

8. No warranties; limitation of liability

The service is provided "as is" without warranties of any kind. Doxa is not a substitute for pastoral care, professional counselling, or medical advice. Responses are generated by an AI model (Anthropic Sonnet) and may occasionally be wrong, awkward, or off-tone despite our best efforts. You are responsible for what you do with the output. To the maximum extent permitted by law, Doxa's liability for any claim arising from use of the service is limited to £100.

9. Suspension & termination

We may suspend or block access (for any IP, key, integration, or downstream product) if we observe abuse, rate-limit circumvention, attempted prompt extraction, ToS violation, or anything else we judge a threat to the service or to end users. We aim to be reasonable; if you think a block is wrong, email abuse@doxa.app.

10. Report abuse or a bug

Two routes:

• Public issue: open an issue in the schema repo.
• Private / sensitive: email abuse@doxa.app for misuse reports, security@doxa.app for security disclosures.

11. Changes to these terms

These terms are versioned by date at the top of this page. Material changes are reflected in a new "Effective" date. Continued use after a change constitutes acceptance.