Get one clear verdict: yes, no, or tie. Route agent actions to deterministic MCP notaries. Add Krafthaus signals only when ranking subjective choices.
Decision API resolves ambiguity. MCP notaries execute deterministic actions. Krafthaus is optional and used only for preference-heavy choices.
Binary verdicts for direct questions, with tie when options are equivalent.
Structured outputs so agents can gate actions before invoking a notary.
Refund, cancel, return, and trial checks. Stateless, policy-sourced, deterministic.
Optional sentiment ranking for best-of-N choices. Notary execution stays deterministic.
A human or agent sends a decision request.
Engine returns yes, no, or tie.
If action is needed, call the matched notary.
When models hedge, decide returns one verdict: yes, no, or tie. No debate loop.
{
\"endpoint\": \"/api/decide\",
\"request\": { \"question\": \"ship this launch today?\", \"mode\": \"single\" },
\"response\": { \"c\": \"yes\", \"v\": \"yes\", \"request_id\": \"abc123\" },
\"field_map\": { \"c\": \"decision_class\", \"v\": \"decision\" }
}
Built for fast choices, clear outputs, and safe defaults.
One output for one input: yes, no, or tie.
If options are equivalent, decide says tie.
Each decision is stateless, so old chat drift does not leak in.
Response format is short and easy to route into workflows.
Finance, legal, and medical prompts are blocked by policy.
Ask a yes/no question or a multi-option prompt.
Receive one verdict: yes, no, or tie.
Execute, escalate, or skip with confidence.
Deterministic MCP servers that give your AI agents clear, auditable answers.
Deterministic refund eligibility checker for US consumer subscriptions. Returns ALLOWED, DENIED, or UNKNOWN based on each vendor's official refund policy window.
Cancellation penalty checker for US consumer subscriptions. Returns FREE_CANCEL, PENALTY, or LOCKED based on each vendor's cancellation terms.
Return eligibility checker for US consumer subscriptions. Returns RETURNABLE, EXPIRED, or NON_RETURNABLE with return type (full refund, prorated, credit) and method.
Run real MCP calls, inspect responses, and keep a scoped trail of recent runs.
Run a request to see output.
Run twice to compare output changes.
| Time | Notary | Mode | Status | Assert | Latency | Req hash | Actions |
|---|
Add decide notaries to any MCP-compatible client in seconds.
{
"mcpServers": {
"refund-decide": {
"url": "https://refund.decide.fyi/api/mcp"
},
"cancel-decide": {
"url": "https://cancel.decide.fyi/api/mcp"
},
"return-decide": {
"url": "https://return.decide.fyi/api/mcp"
},
"trial-decide": {
"url": "https://trial.decide.fyi/api/mcp"
}
}
}
curl -X POST https://refund.decide.fyi/api/mcp \ -H "Content-Type: application/json" \ -d '{ "jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": { "name": "refund_eligibility", "arguments": { "vendor": "adobe", "days_since_purchase": 12, "region": "US", "plan": "individual" } } }'
curl -X POST https://refund.decide.fyi/api/v1/refund/eligibility \ -H "Content-Type: application/json" \ -d '{"vendor":"spotify","days_since_purchase":5,"region":"US","plan":"individual"}'
Start free, then upgrade when your team needs higher volume and workflows.
Support is async and best-effort on Starter and Pro. No SLA outside Enterprise contracts.
No API keys. No databases. No side effects.
Same input, same output. Every time. Fully auditable decisions your agent can cite.
No sessions, no tokens, no stored data. Pure function over HTTP.
JSON-RPC 2.0 over HTTP POST. Works with Claude Desktop, Cursor, and any MCP client.
Adobe, Netflix, Spotify, Microsoft 365, and 60 more. Updated daily via automated checks.
No API keys, no auth, no sign-up. Point your agent at the endpoint and go.
Full rules, policy sources, and server code on GitHub. Verify everything.
Last updated: February 2026
decide provides consumer decision tools for humans and AI agents, including MCP notaries and a live demo with audit trail.
What we process: request inputs and outputs needed to run your call. Local demo history, saved cases, and assertions are stored in your browser unless you use authenticated account features.
What we don't do: we do not sell personal data and we do not use your prompts to train public models.
Operational logs: we may keep service logs (for uptime, abuse prevention, and debugging) with limited metadata such as timestamp, route, status, latency, and hashed request identifiers.
Questions? Email decidefyi@gmail.com.
Last updated: February 2026
Service scope: decide offers decision support tools (human yes/no flows, MCP notaries, live demo, and audit trail).
No professional advice: outputs are informational and may be incomplete. They are not financial, legal, or medical advice.
Safety policy: certain high-risk prompts may be blocked (for example financial/legal/medical asks) and the UI will return an explanatory message.
Your responsibility: you are responsible for downstream decisions and integrations built on top of decide responses.
Availability: features may change, and uptime is best-effort while the platform evolves.
What you can use today
• MCP notaries: refund, cancel, return, and trial
• REST endpoints for the same deterministic policy checks
• Live demo + audit trail for testing, assertions, diffs, and shareable snapshots
Best for
• Agent workflows that need deterministic policy answers
• QA and regression checks with assertion templates
• Teams that need reproducible run history and troubleshooting context
Integration help: contact decidefyi@gmail.com.
Answer 4 quick questions and I'll reply with a Pro pilot link + suggested setup. This keeps support async and lightweight.
For product feedback, integrations, or partnership requests:
• X: @decidefyi
• Email: decidefyi@gmail.com
If you're sharing a demo issue, include your run snapshot link so we can debug quickly.