See every step your AI agents take.
Self-hosted monitoring for AI agents and LLM apps. One container and one Postgres, running in your own cloud. Your prompts and data never leave it.
Preview design · sample data
Know what your agents did, what it cost, and where it broke.
The features planned for the first release.
Step-by-step traces
Every prompt, model call, tool call and answer in one timeline, so you can see exactly why an agent said what it said.
Cost and tokens
Cost per conversation, per user and per model, so a runaway loop or a pricey prompt shows up the same day.
Latency and errors
Slow steps, failed tool calls and model errors, with the full trace one click away.
Alerts
Error rate, cost or latency over your limit? Get told on email, Teams, Slack or WhatsApp.
Personal data masking
Phone numbers, emails, Aadhaar and PAN masked before they are stored. Retention you control.
Open standards
Uses OpenTelemetry and the GenAI conventions. Works with OpenAI, Gemini, Claude, LangChain, LlamaIndex and more. No lock-in.
Ten minutes to your first trace.
Run it in your cloud
One Docker image plus PostgreSQL. Ready templates for Azure Container Apps, AWS and Google Cloud Run, or plain Docker Compose.
Point OpenTelemetry at it
Use the standard OpenTelemetry libraries you may already have. No custom SDK to learn.
See every run
Open the dashboard: traces, costs, latency and errors for every agent and every user.
Planned quick start (subject to change)
# 1. Start NexAITrace and Postgres docker compose up -d # 2. Send traces from your Python agent pip install opentelemetry-sdk opentelemetry-exporter-otlp \ opentelemetry-instrumentation-openai export OTEL_EXPORTER_OTLP_ENDPOINT=https://trace.your-company.internal export OTEL_EXPORTER_OTLP_HEADERS="x-api-key=your-project-key" # 3. Run your agent as usual: every call is traced opentelemetry-instrument python agent.py
Built for teams who want it simple, in their own cloud.
How we plan to compare with typical self-hosted LLM monitoring setups.
| Typical self-hosted setup | NexAITrace | |
|---|---|---|
| What you run | Several services, often with ClickHouse, Redis and object storage | 1 container + PostgreSQL |
| Where data lives | Your cloud, or the vendor's cloud for hosted plans | Always your cloud |
| Pricing | Enterprise tiers usually in US dollars | Free community edition · licences in INR with GST invoices |
| Alerts | Email or Slack | Email, Teams, Slack and WhatsApp |
| Support | Community or overseas | India time zone, direct |
Run AI agents in production? Help shape NexAITrace.
We're inviting a small group of teams to try NexAITrace free in their own cloud and tell us what they need. Tell us which cloud you use and roughly how many AI calls you make a day.