Frequently asked questions

You need your own Anthropic API key and an MCP-compatible AI client that can send a custom request header. Claude Code, Codex CLI, and Gemini CLI are supported. Other clients can connect if they support streamable HTTP and custom headers.

Follow the connection guide, then ask about your car, a symptom, or a build goal in plain language.

Add the rx7db MCP endpoint and pass your key in the x-api-key header. You'll find the full command in the connection guide.

rx7db researches each request before it answers, so a detailed question can take a few minutes. The server stops a run after ten minutes. Set your client's tool timeout above that limit. In Claude Code, MCP_TOOL_TIMEOUT controls tool calls in milliseconds, while MCP_TIMEOUT only controls MCP server startup.

We don't charge for access to rx7db right now. You supply your own Anthropic API key, and Anthropic bills its usage to your account. We don't ask for payment details or charge a separate subscription.

Most tool calls cost about US$0.05–$0.50. The final cost depends on the question, the amount of research involved, and the model available to your Anthropic account. Your Anthropic usage dashboard shows the exact amount.

Yes. You can manage spend limits, usage alerts, and key access in your Anthropic account. Focused questions usually cost less. Include the relevant vehicle details up front to avoid extra follow-up calls.

rx7db can troubleshoot symptoms, plan repairs and installs, explain wiring and vacuum routing, answer cooling and tuning questions, compare turbo combinations, check wheel fitment, locate parts, and plan a build. Use the general ask tool for anything that doesn't fit one of those categories.

Yes. rx7db researches factory service procedures, community guides, and up-to-date search results. When a diagram helps, it links to the image so you don't have to rely on a text-only explanation.

Yes. Add a vehicle_config with your year, market, ECU, turbo setup, and supporting modifications, or save one with set_vehicle. Each answer also includes a session_id. Pass it to another rx7db tool to continue the same conversation. Sessions expire after seven days of inactivity.

Like all AI software, rx7db can make mistakes, and the community guides it reads on the internet can be wrong. Always check critical details against the right service information for your specific car. Incorrect tuning, fuel, electrical, or mechanical work can damage the engine or put people at risk. You are responsible for deciding how to use each answer, and we make no guarantees that any answer is fit for any particular use case.

No. rx7db has not trained and will not train a custom AI model on your content or anyone else's. It uses existing information sourced from Claude models and web research tools as source materials to answer your questions. This information may go through caching and transformation steps to create better answers.

rx7db does not know or control what material Anthropic used to train Claude. Questions or requests about Claude's underlying models need to go to Anthropic through its Privacy Center.

rx7db researches any information currently publically available on the internet, including owner communities, guides, publications, documents, instructions, and any content assimilated by existing AI models from Claude and/or OpenAI. The individual sources vary with the question, and answers may include citations, source links, or links to useful diagrams from these sources.

No. Third-party materials remain the property of their respective owners. rx7db uses them as research sources and generally summarizes the relevant information in its own words, with citations or links when useful. See Copyrights in the Terms of Service for more detail.

Sometimes, but there is no single place that controls every result. If the information comes from a website, ask the site owner to remove or correct the source. You can then ask Google or another search provider to update its index. Removing a result from Google Search does not remove the source page from the internet.

If an answer reflects information in Claude's underlying model, only Anthropic can review that request. rx7db cannot modify Claude. Copies and alternative sources may also exist, so these steps cannot guarantee that the subject will never appear in a result.

In short: All answers are read from the internet. We don't own or control the internet. If the information is there, and Claude can find it, odds are it'll appear here. We're just really good at finding it when it matters.

No. The key is supplied in the request header and passed only to the internal agent for that call. We don't store or log the raw key. We only keep a short SHA-256 fingerprint so a conversation session can be restricted to the key that created it.

• Follow the setup guide when adding the key to your assistant.
• For private-beta enrollment, hash the key locally and send us only the 64-character SHA-256 digest.
• Never share your full API key in a prompt, source file, screenshot, forum post, or chat message.
• Use a dedicated key with the spend controls you want in your Anthropic account.

Revoke or rotate the key in your Anthropic account right away, create a replacement, and update your MCP client configuration. Any existing rx7db session tied to the old key will no longer work with the new one, which is expected.

No. rx7db runs its research tools through Anthropic, so you need an Anthropic API key. OpenAI doesn't allow you to send your API key in a header, so we aren't able to use their models yet.

We may add OAuth and direct billing later so you won't need to configure a provider key in your client. This will also let you use the service through web or mobile based AI clients and applications.