Trust scores for every AI response. Local-first. Model-agnostic.
TrustLayer sits between your app and any LLM. It verifies outputs, catches hallucinations, tracks token costs, and gives every response a trust score -- all running locally on your machine.
Works with Ollama, Claude, GPT, Gemini out of the box.
- Trust Scoring -- confidence ratings on every AI response, so you know what to double-check
- Hallucination Detection -- flags claims that contradict your source docs or prior context
- Cost Tracking -- per-request and cumulative token spend across all providers
- Local Proxy -- routes requests through a local server; your data never hits a third-party logging service
- Multi-Model Support -- swap between Ollama, OpenAI, Anthropic, and Google models with one config change
- Dashboard -- browser-based UI to inspect responses, trust history, and cost breakdowns
pip install trustlayer-aifrom trustlayer import TrustLayer
tl = TrustLayer()
result = tl.verify(
prompt="What is the capital of France?",
model="ollama/llama3"
)
print(result.answer) # "Paris"
print(result.trust_score) # 0.97
print(result.cost_usd) # 0.0 (local model)Screenshots coming soon. Placeholders below.
| Dashboard | Trust History |
|---|---|
 |
 |
| Feature | TrustLayer | Open WebUI | LiteLLM | PrivateGPT |
|---|---|---|---|---|
| Trust scoring per response | Yes | -- | -- | -- |
| Hallucination detection | Yes | -- | -- | Partial |
| Cost tracking | Yes | -- | Yes | -- |
| Local-first proxy | Yes | Yes | Yes | Yes |
| Multi-model routing | Yes | Yes | Yes | -- |
| Python SDK | Yes | -- | Yes | -- |
- Developers building AI features who need to know when the model is wrong
- AI/ML teams tracking spend and reliability across providers
- Self-hosters who want full control -- no cloud dependency, no telemetry
Contributions welcome. Open an issue first for anything non-trivial.
git clone https://github.com/acunningham-ship-it/trustlayer.git
cd trustlayer
pip install -e ".[dev]"MIT