Ask one question in plain English — AI agents query metrics, traces, and topology together, then tell you what broke.
Live Demo | Documentation | 中文 | Community
- 🤖 AI-native, not a bolt-on chat box — LLM queries traces, metrics, topology, and alerts directly from real data
- 🧠 Multi-agent collaboration — AI Brain orchestrates query and inspection experts; complex tasks run in parallel
- 🎯 AI application observability (Roadmap) — LLM call chains · token analytics · agent topology · skill/tool/model tracing
- ⚡ eBPF APM (Roadmap) — kernel-level, non-intrusive collection without code changes
- 📊 OpenTelemetry APM foundation — OTLP ingestion with troubleshooting, tracing, service metrics, and topology
- 🚨 Alerting loop — threshold and change detection, scheduled evaluation, alert event history
- 🔧 Skill + Tool extensibility — override built-in skills, add custom digital experts without touching core code
- 🔌 MCP both ways — expose platform capabilities to Cursor / Claude; ingest external MCPs (Prometheus, SkyWalking, etc.)
- 🐳 Minimal 3-component stack — Ingest + Doris + Web; one Docker / K8s command, no middleware sprawl
- 🌐 Bring your own model — OpenAI-compatible + Anthropic Messages; Kimi, DeepSeek, GLM, Bailian, Qianfan, Ollama, and more
AI multi-agent troubleshooting · Service health · Call graph topology — English UI
| DataBuff | SigNoz | Datadog | |
|---|---|---|---|
| AI multi-agent root cause | ✅ Built-in | ❌ | Add-on |
| OpenTelemetry native | ✅ | ✅ | Partial |
| Self-host · 5 min Docker | ✅ curl | bash |
✅ | ❌ SaaS |
| Distributed tracing + topology | ✅ | ✅ | ✅ |
| Open source | ✅ | ✅ | ❌ |
One-liner: Open-source APM where an AI brain dispatches metric, trace, and inspection experts — not another chat box on top of dashboards.
AI Analysis
Natural language query · Metrics and traces in plain language |
Multi-agent collaboration · Evidence synthesis and conclusions |
APM Observability
Service list · Traffic-light status for anomalies |
Global topology · Auto-generated call graph |
Service detail · Metric trends and instances |
Service flow · Upstream and downstream dependencies |
⚡ From running the install command to demo apps reporting data and showing traces and topology, you can see results in about 5 minutes.
Requires docker and docker-compose. The install script auto-detects amd64/arm64 and downloads the matching image bundle.
1. Install Platform
curl -fsSL https://databuff.ai/databuff/ai-apm-install.sh | bash2. Install Demo App (optional)
curl -fsSL https://databuff.ai/databuff/ai-apm-demo-install.sh | bashOffline Install
When the registry is unreachable, download the bundle for your architecture and install on the target machine. Pick a version on the install page under Docker → Offline Install, or use:
https://openocta.com/pkg/databuff/<version>/offline/databuff-ai-apm-offline-<version>-<arch>.tar.gz
tar -zxvf databuff-ai-apm-offline-<version>-<arch>.tar.gz
cd databuff-ai-apm-offline-<version>-<arch>
# Install platform
sudo ./install.shKubernetes
Requires kubectl and a working Kubernetes cluster. The script installs the platform via K8s manifests.
1. Install Platform
curl -fsSL https://databuff.ai/databuff/ai-apm-k8s-install.sh | bash2. Install Demo App (optional)
curl -fsSL https://databuff.ai/databuff/ai-apm-demo-k8s-install.sh | bashOffline image download
If the install commands above cannot pull images due to network issues, run the following to download an offline image bundle and load it onto the node.
curl -fsSL https://databuff.ai/databuff/ai-apm-k8s-download-images.sh | bash
Open http://YOUR_HOST:27403 · Default login admin / Databuff@123 · Add your API key in model settings to enable AI
Scan the QR code to join the Databuff open-source community on WeChat
