AutoSRE

🤖 The AI SRE that investigates incidents like your best on-call engineer — but faster.

Quick Start • Features • How It Works • Integrations • Docs

45-minute investigations → 5 minutes. Autonomous triage. Evidence-based RCA. Human-in-the-loop for safety.

⚡ Quick Start

# Install
pip install autosre-ai

# Configure (interactive setup)
autosre config init

# Investigate your first incident
autosre investigate "checkout service 500 errors" --service checkout-service

Or with Docker:

docker run -it --rm -v ~/.autosre:/root/.autosre ghcr.io/autosre-ai/autosre investigate "high latency on api-gateway"

That's it. No Neo4j. No Postgres. No infrastructure. Just pip install and go.

✨ Features

🔍 Autonomous Investigation

Multi-agent investigation that works like your best SRE: triage → contain → investigate → resolve → learn.

$ autosre investigate "payment failures spiking"

[Triage] Confirmed: payment-service 5xx rate at 12% (normally <0.1%)
[Scope] Affected: checkout-service, order-service (downstream)
[Hypothesis] Testing: Recent deployment of payment-service v2.3.1
[Evidence] Deployment at 14:02, errors started 14:05 ✓
[Root Cause] payment-service v2.3.1 introduced null pointer in retry logic
[Recommendation] Rollback to v2.3.0 (requires approval)

🧠 Episodic Memory

Learns from every investigation. Recalls similar incidents. Gets smarter over time.

$ autosre memory search "database timeout"

Found 3 similar incidents:
├── inv_abc123: PostgreSQL connection pool exhaustion (resolved in 8m)
├── inv_def456: Slow query blocking connections (resolved in 12m)
└── inv_ghi789: Network partition to RDS (resolved in 23m)

📊 SLO-Driven Operations

Error budgets, multi-window burn rates, deployment gating — all built-in.

$ autosre slo status --service checkout-service

checkout-service SLO Status
├── Availability: 99.92% (target: 99.9%) ✓
├── Latency p99: 245ms (target: 300ms) ✓
├── Error Budget: 72% remaining
│   ├── 1h burn rate: 0.8x
│   ├── 6h burn rate: 1.2x
│   └── 24h burn rate: 0.9x
└── Deploys: ALLOWED

🛡️ AI Safety Built-In

Every decision has confidence scores. Critical actions require human approval. Full audit trails.

• Hypothesis-driven reasoning with falsifiable criteria
• Confidence scoring (0.0-1.0) on every decision
• Human-in-the-loop for remediation actions
• AI error budgets tracking accuracy over time

🔧 Extensible Skills System

Modular investigation skills: Kubernetes, metrics, logs, traces, infrastructure.

skills/
├── kubernetes/         # Pod states, deployments, events
├── metrics-analysis/   # Prometheus, Datadog, Grafana
├── log-analysis/       # Pattern matching, anomaly detection
├── traces/             # Distributed tracing analysis
├── infrastructure/     # AWS, GCP resource checks
└── investigation/      # Methodology and hypothesis testing

📝 Automated Postmortems

Blameless postmortems with auto-generated timelines, metrics snapshots, and action items.

🎯 How It Works

┌────────────────────────────────────────────────────────────────┐
│                     autosre investigate                         │
└────────────────────────────┬───────────────────────────────────┘
                             │
                    ┌────────▼────────┐
                    │   Orchestrator   │
                    │   (LangGraph)    │
                    └────────┬────────┘
                             │
         ┌───────────┬───────┴───────┬───────────┐
         │           │               │           │
    ┌────▼────┐ ┌────▼────┐   ┌─────▼────┐ ┌────▼────┐
    │ Memory  │ │Topology │   │ Planner  │ │  LLM    │
    │(SQLite) │ │ (YAML)  │   │  Agent   │ │ Router  │
    └─────────┘ └─────────┘   └────┬─────┘ └─────────┘
                                   │
               ┌─────────┬─────────┼─────────┬─────────┐
               │         │         │         │         │
          ┌────▼───┐┌────▼───┐┌────▼───┐┌────▼───┐┌────▼───┐
          │  K8s   ││Metrics ││  Logs  ││ Traces ││ Infra  │
          │Subagent││Subagent││Subagent││Subagent││Subagent│
          └────┬───┘└────┬───┘└────┬───┘└────┬───┘└────┬───┘
               └─────────┴─────────┴─────────┴─────────┘
                                   │
                    ┌──────────────┴──────────────┐
                    │         Synthesizer         │
                    │   (Evidence → Root Cause)   │
                    └──────────────┬──────────────┘
                                   │
                    ┌──────────────▼──────────────┐
                    │     Writeup & Actions       │
                    │  (Postmortem, Remediation)  │
                    └─────────────────────────────┘

Key Concepts:

🔌 Integrations

📈 Why AutoSRE?

Test Results: 1,053 tests passing | 25+ investigation scenarios validated

🏗️ Production Deployment

For production deployments with persistent storage and multiple services, see the Docker Deployment Guide.

# Clone and setup
git clone https://github.com/autosre-ai/autosre.git
cd autosre
make setup

# Configure secrets
cp .env.example .env
vim .env  # Add your API keys

# Start all services
make dev

# Verify health
make health

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

# Development setup
git clone https://github.com/autosre-ai/autosre.git
cd autosre
pip install -e ".[dev]"
pytest  # Run the test suite

Areas we need help:

• 🔌 New integrations (Elastic, Splunk, New Relic)
• 📊 Investigation scenarios for evaluation
• 📚 Documentation and examples
• 🐛 Bug reports and fixes

📄 License

Apache 2.0 — See LICENSE for details.

Built by SREs, for SREs.
_{Tired of 3am pages? Let AutoSRE handle the first 5 minutes.}

⭐ Star us on GitHub • 💬 Join Discord • 🐦 Follow on Twitter