Open Source · Local-first · Privacy by default

You bring the AI.
We bring the trust.

The universal trust layer for every AI tool you use. Verify outputs, track costs across all providers, and compare models on your actual tasks — running entirely on your machine.

Get started free See how it works →
$ pip install trustlayer && trustlayer server copy
No account. No config. Works in 30 seconds.
8
Core features
0
Required accounts
100%
Local by default
MIT
License
Verification Engine

Trust scores on every output

TrustLayer analyzes each AI response for hallucinations, overconfidence, and unverifiable claims — then gives you a simple 0–100 score with specific flags.

AI Response — Overconfident claim real output from /api/verify

"There are definitely 100% more jobs created by AI than lost. Every single economist agrees. The truth is AI will certainly create infinite wealth for everyone by 2025."

85
High trust — 1 concern flagged
⚠ High confidence language with no hedging — may be overconfident
AI Response — Cited claim real output from /api/verify

"Research suggests that regular exercise may reduce heart disease risk by approximately 30–35%, though individual results vary. According to WHO guidelines, 150 minutes per week of moderate activity is the baseline."

88
High trust — 2 sources cited
✓ Hedged language used ✓ 2 citations detected ⚠ Verify specific statistic independently
TrustLayer Dashboard showing trust scores and verification results

The TrustLayer dashboard — trust scores at a glance

What's included

Everything you need to trust AI

Eight core features that make TrustLayer the last AI tool you need to add to your workflow.

⚡

Universal Connector

Plug in any AI: Ollama (auto-detected), Claude, GPT-4, Gemini, Aider. One unified interface for all of them. Switch between providers seamlessly.

Ollama · Anthropic · OpenAI · Google
TrustLayer provider connectors
🛡️

Verification Engine

Every AI output gets a 0–100 trust score. Hallucination detection, overconfidence flags, and claim-by-claim source attribution.

Trust Score · Hallucination detection
🧠

Personal Learning

Remembers how you work across sessions. Learns your writing style, common tasks, and preferences. Your 500th session is better than your 1st.

Stored locally · No cloud sync
💰

Cost Tracker

Real-time spending dashboard across all AI providers. Budget alerts and per-task cost optimization tips to cut your bill.

Budget alerts · Monthly reports
⚖️

Model Comparison

Test your actual tasks against multiple models side-by-side. Personal benchmarks — not generic leaderboard scores that don't match your workflow.

Side-by-side · Speed · Accuracy · Cost
📚

Offline Knowledge Base

Drag in your docs, notes, PDFs, and code repos. Everything indexed locally. AI uses your context when answering. Works fully offline with Ollama.

Local indexing · Works offline
📜

History

Complete log of every AI interaction. Trust scores, costs, latency, and full prompt/response pairs — searchable and filterable by provider.

Searchable · Filterable · Full audit trail
🎭

Adaptive Personality

Honest and humble for factual queries. Creative and opinionated for brainstorming. Invisible for routine tasks. Adapts automatically to context.

Context-aware · Auto-adapts
TrustLayer interaction history and audit log

Complete interaction history — every AI call logged and searchable

How it works

Simple to start. Powerful over time.

TrustLayer sits between you and every AI you use. You don't change your workflow — you just gain visibility and trust.

1

Connect your AIs

Ollama is auto-detected. Add API keys for Claude, GPT-4, or Gemini via the Settings page or environment variables. No account required for TrustLayer itself.

2

Every output gets verified

As you use AI, TrustLayer silently scores each response. Trust scores, hallucination flags, and source attribution appear inline — without slowing you down.

3

It learns your patterns

TrustLayer remembers your sessions, preferences, and writing style. Over time, it surfaces better suggestions, cheaper alternatives, and personalized tips.

Quick Start

Simple by design

Works with everything you already have. Zero config. One command to start.

# Install TrustLayer
pip install git+https://github.com/acunningham-ship-it/trustlayer.git

# Start the server + web UI
trustlayer server
# → Auto-detects Ollama
# → Opens http://localhost:8000

# Add your API keys (optional)
export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=sk-...
# Verify any AI output
trustlayer verify "The earth is 4.5 billion years old."
# → Trust Score: 94/100 (HIGH) — No concerns

# Ask any connected AI
trustlayer ask "Summarize this codebase" --provider ollama --model llama3.2

# Compare responses from multiple models
trustlayer compare "Write unit tests for this function"

# Check your spending across all providers
trustlayer costs

# Detect available AI tools on your machine
trustlayer detect
# Verify content via REST API
curl -X POST http://localhost:8000/api/verify \
  -H "Content-Type: application/json" \
  -d '{"content": "Your AI output here"}'

# Actual response:
{
  "trust_score": 85,
  "trust_label": "high",
  "summary": "85% trusted. 1 concern(s) flagged. 0 source(s) cited.",
  "issues": ["High confidence language with no hedging"],
  "hallucination_score": 15
}

# Compare models on the same prompt
curl -X POST http://localhost:8000/api/compare \
  -d '{"prompt": "Explain quantum entanglement"}'
import requests

# Verify any AI output
response = requests.post("http://localhost:8000/api/verify", json={
    "content": "AI output to verify"
})
result = response.json()
print(f"Trust: {result['trust_score']}/100 — {result['trust_label']}")

# Compare providers side by side
comparison = requests.post("http://localhost:8000/api/compare", json={
    "prompt": "Your task here",
    "providers": ["ollama", "anthropic", "openai"]
}).json()
Why TrustLayer

What you're missing without it

No other tool gives you verification, cost tracking, personal learning, and offline support — all in one place that runs locally.

Capability TrustLayer Open WebUI Individual AI apps
Output verification & trust scores ✓ ✗ ✗
Cross-provider cost tracking ✓ ✗ ✗
Personal learning across sessions ✓ ✗ ~
Side-by-side model comparison ✓ ~ ✗
Offline local knowledge base ✓ ~ ✗
Complete interaction history & audit log ✓ ✗ ✗
100% local — no cloud required ✓ ✓ ✗
Open source, MIT license ✓ ✓ ✗
The problems we solve

Real pain points driving real developers away from AI

These aren't hypothetical — they're the top complaints across Reddit, HN, and X right now.

Using 3+ AI tools daily but no way to know which output is actually correct when they disagree.

The trust problem

$200/month across Claude, OpenAI, and Gemini — with zero visibility into what you actually spent it on.

The cost problem

AI presents hallucinated facts with the same confidence as verified ones. No way to tell the difference until it's too late.

The hallucination problem

Sensitive data sent to cloud AI servers with no control over storage, training, or access. Local alternatives exist but lack quality.

The privacy problem

Get started

Stop guessing.
Start trusting.

TrustLayer is free, open source, and runs entirely on your machine. Your data never leaves your computer. No account required.

View on GitHub Read the docs