ThinkWatch Core#

ThinkWatch Core is the shared core of the ThinkWatch AI API gateways: routing, forwarding, observability, cost accounting, and a set of data-plane guards. It is used by the desktop app, ThinkWatch Lite, and by the server edition.

Scope of Core#

Core is a set of Rust crates, not an installable application. For a runnable gateway, bin/twcore is a complete, self-contained gateway binary; Quick start with twcore describes how to use it.

Both editions depend on these crates, so a change in Core affects both.

What it does#

When a client such as Claude Code or Codex is pointed at a local port, Core provides the following:

• Rule-based routing to different upstreams. Conditions include the model name, the client, the context length, and whether tools are present; actions include switching upstream, rewriting parameters, and rejecting the request.
• Mid-stream failover. Before the first byte is sent, the gateway can switch upstreams transparently. After streaming has started, it reports the failure.
• Cost visibility. Token usage and cache hits are priced against a snapshot table. Usage that cannot be priced is labelled unknown rather than assigned a fabricated figure.
• Outbound redaction. Secrets in a request are replaced with placeholders before they reach a relay, and restored when the model echoes them back.
• Inbound inspection. Tool calls returned by an upstream are checked against a rule set, and a dangerous call can be terminated mid-frame.

Further reading#

• Quick start with twcore: writing, checking, and serving a config.
• Crate layers: how the crates are organised, and which layers are shared.
• Development and tests: tests, the smoke script, and the rules every change must follow.

License#

ThinkWatch Core is licensed under the MIT License. The source code is available on GitHub.