MCP (Model Context Protocol) lets an AI assistant use a defined set of tools supplied by another service. A developer can expose a search function or a vehicle-state reader through MCP, then let a compatible assistant call it within the account’s permissions.
Two independent projects maintained under GAL1LEO illustrate this approach: a WhatsApp bridge and a Tesla Fleet API integration. Both run on infrastructure supplied by the operator.
WhatsApp: working with conversations
The WhatsApp bridge provides message search, access to conversations and attachments, user-requested sending, event subscriptions and optional local voice-note transcription. A simple use case is asking an assistant to summarise selected messages and identify questions that still need a reply.
The connection uses WAHA/GOWS, an unofficial integration for personal WhatsApp accounts. A linked-device archive is a working source of messages; it does not guarantee a complete copy of the phone’s history. Sending and live account access start disabled.
The basic path is:
dedicated WhatsApp account/session → WAHA/GOWS → bridge and local database → HTTPS/OAuth → MCP client
WhatsApp source and installation guide
Tesla: reading observations and requesting actions
The Tesla integration uses the Tesla Fleet API. It exposes vehicle and charging information, stored telemetry/history, an owner dashboard, configurable rules and event subscriptions. An assistant could explain a charging observation or prepare a report from data the service has collected.
Physical commands require explicit intent and confirmation, with availability determined by permissions, vehicle hardware, firmware and the operation’s requirements. Events do not execute vehicle commands. Reading stored state does not implicitly wake the car.
Tesla source and installation guide
Using the projects with Codex, ChatGPT plugins and Dots
The repository includes skills and client-packaging instructions. The operator first deploys a backend, configures authentication and then connects it to the chosen MCP client. ChatGPT web packaging also requires an app registration belonging to the destination account.
For Codex and ChatGPT, the useful pattern is to combine conversation with explicitly available tools: retrieve selected data, explain it, draft a response or request an authorised action. With Dots, possible extensions include recurring conversation summaries or charging reports, where plugin access and event support are available. Compatibility and native notification delivery need validation in each host.
What a deployment requires
These are source distributions for developers comfortable operating a server. Setup requires Linux, Python, an HTTPS domain, OAuth/OIDC configuration and the operator’s own accounts and credentials. Tesla additionally requires a Fleet API application registration; telemetry and signed commands have their own setup steps.
Original GAL1LEO code is licensed under Apache-2.0, with separate notices for third-party components. There is no licence fee for the original code; hosting and provider/API usage remain separate costs. The projects are independent of WhatsApp, Tesla and OpenAI.
The public documentation separates synthetic test results from live-account and device acceptance. Installation feedback and reproducible technical issues are welcome through the repository.
Generated by an AI assistant using the public project documentation.
GAL1LEO



