Data Use

Last updated: July 2026

This page explains how Zephex uses data that flows through the Service — tool calls, request metadata, dashboard usage, and team rollout responsibilities.

What runs where

  • MCP Endpoint (zephex.dev/mcp) — your client connects here (Vercel).
  • MCP Server (Railway) — tool execution; inputs/outputs stay in memory for the request only.
  • Database (Supabase Postgres) — account data, API key hashes, usage counters, billing metadata.
  • Authentication — Supabase Auth (dashboard), Auth0 (MCP connector OAuth only), Bearer API keys for tool calls.

Uptime: zephex.dev/status. Technical detail: Data Handling (docs). Team rollout: Team policy.

We do not train AI models on your content

Zephex does not use prompts, tool inputs, tool outputs, or repository content accessed through tools to train, fine-tune, or evaluate AI models. This applies to all plans without exception.

  • We do not sell your content to third parties.
  • We do not use your content for advertising.

Teams and organizations

If you roll Zephex out to a team, the account owner remains responsible for every API key created under that account. Each developer should use named keys per environment; shared keys in chat or public repos violate our acceptable-use rules.

  • The account owner is responsible for all API key usage under their organization, including contractors.
  • API keys must not be embedded in client-side code, public repositories, or shared chat channels.
  • Each team member should use their own named key where possible so usage and revocation are attributable.
  • Pro and Max plans may restrict keys to specific tools; calls outside the allowlist are rejected.
  • Users must keep at least four of ten MCP tools enabled in editor configuration.
  • Exceeding plan quotas or abusing rate limits may result in throttling (HTTP 429) or suspension.
  • Enterprise volume or custom limits: contact support@zephex.dev.

The 10 MCP tools — what each processes

Each tool receives only the arguments you send (paths, queries, URLs). Outputs return to your editor and are not persisted by Zephex.

ToolPurposeTypical team use
get_project_contextProject ContextFirst on a new repo before guessing install/test commands. Follow related_topics for deeper slices.
read_codeCode ReaderKnown symbol or files[] from find_code. Outline before 300+ line files. Scan/smell on 3–8 focus files.
find_codeCode SearchUnknown location; pasted line; rename before editing; multi-keyword topic hunt.
explain_architectureArchitecture AnalyzerBefore cross-cutting edits — how does this feature/subsystem span modules?
Zephex_dev_infoZephex Dev InfoGeneric playbook after repo tools if local context is insufficient.
audit_headersSite AuditUser pastes live URL — post-deploy, security, framework, secrets, is it up.
check_testTest PulseAfter code changes, before push, or when user asks if tests pass — not to find which files to edit.
check_packagePackage CheckBefore npm/pip/cargo install or version bump. User asks "is X safe?" or "how do I upgrade X?"
keep_thinkingKeep ThinkingHigh-blast-radius edit (auth, billing, schema); debugging after 2+ failed attempts; task spans 3+ files.
project_memoryProject MemoryUser says remember/recall; before unfamiliar subsystem; after non-obvious decision.

Tool reference: docs/tools

What is stored vs. what is not

DataStored?Duration
Source code in tool callsNo
Tool input argumentsNo
Tool output / response bodyNo
AI chat / promptsNo
Tool name per callYes90 days
Timestamp + success/error + latencyYes90 days
API key hash (HMAC-SHA256 + salt)YesUntil revoked
Hashed IP addressYes90 days

Billing and quota

  • One successful MCP tool call = one request against the account monthly quota.
  • Failed tool calls do not bill.
  • Cached Test Pulse readbacks (failures, summary, test-last) are 0 credits when session TTL is valid.
  • Meta commands (help, doctor, usage, learn) do not consume quota.

Plans: Pricing tiers

What data may be processed (ephemeral)

Depending on what you do, the Service may process in memory:

  • Tool call inputs/outputs (URLs, text, or code snippets you provide).
  • File paths and search patterns sent to find_code or read_code.
  • Public GitHub URLs when using hosted path modes.

None of the above is written to disk or retained after the HTTP response completes.

How we use metadata

  • Provide the Service: route requests, return responses, show usage in the dashboard.
  • Security: detect abuse, enforce rate limits, investigate suspicious activity.
  • Billing: measure usage and reconcile billing events.
  • Reliability: debug errors, monitor uptime, and improve performance.

Logging and retention

We log the minimum required for security and billing: tool name, timestamp, status, and latency — not tool arguments or response bodies. Usage events roll off after 90 days. Account deletion removes profile and key data within 30 days (billing records may be anonymized per tax law).

Your choices

  • Revoke or rotate API keys at any time in the dashboard.
  • Decline optional analytics cookies (EU) via the site cookie banner.
  • Request export or deletion of account data: support@zephex.dev

Contact

Data questions: support@zephex.dev