Skip to content

DISCLAIMER.md

Latest commit

 

History

History
53 lines (39 loc) · 3.08 KB

DISCLAIMER.md

File metadata and controls

53 lines (39 loc) · 3.08 KB
Raw

Disclaimer

TangleBrain is a local-first, config-driven router across OpenAI-compatible backends you own. Out of the box it ships with a single active backend — a free local model server — and routes only there. Every other backend is opt-in: you enable it by editing your own roster config. This document describes what you take responsibility for when you do.

No warranty

TangleBrain is provided under the MIT License "as is", without warranty of any kind. You are responsible for how you configure and run it, including which backends you connect and the costs and obligations those backends carry.

Subscription / authenticated-CLI adapters are your responsibility

TangleBrain can drive command-line tools you already have installed and logged in (for example Claude Code, Codex, or the Gemini CLI) through its generic cli adapter. These entries ship commented out — they are examples, not defaults. If you uncomment one, you choose to point TangleBrain at that tool.

Using those tools through TangleBrain is your responsibility under each provider's Terms of Service. TangleBrain does not bundle, authenticate, or speak for any provider; it invokes whatever command you configure with the credentials already present on your machine. Before enabling a cli adapter, confirm that automating that tool the way you intend is permitted by the provider whose service it uses. If a provider's terms prohibit a given use, don't configure TangleBrain to do it.

The cli adapter is intentionally generic: it runs a configured command as a subprocess (never via a shell) and parses its output. It has no special knowledge of any particular vendor, and these docs name specific CLIs only as configuration examples.

Paid-API tier is bring-your-own-key and off by default

The paid-API tier costs real money, so it is disabled by default and gated by two independent switches that must both be on for a paid backend to be reachable:

  1. the global api_billing_enabled flag in config/settings.yaml (ships false), and
  2. the per-entry enabled flag on the roster entry.

The durable rule is: no paid billing without the explicit toggle. A tier: api entry parses and is inspectable at all times but is never routable until you turn both on.

When you do enable it, you bring your own key. TangleBrain never holds a raw secret — a key_ref points at an environment variable or a 0600 file, resolved only at call time. TangleBrain records a budget_usd_month for visibility but does not enforce spend; cap your spend at the source (a budget-capped or scoped key, or a gateway that enforces a hard limit). You are responsible for any charges your key incurs.

Costs, rate limits, and data

Whatever backend you route to, you own the relationship with that backend — its costs, its rate limits, its data-handling, and its terms. TangleBrain passes your prompts to the backend you configured and returns the result; it adds no telemetry and sends nothing anywhere you didn't point it. Review each backend's data and privacy terms before sending it anything sensitive.