# Ongrid

> Self-hostable ops AI agent that investigates infrastructure incidents, finds root causes, and acts from Slack, Telegram and other chat channels.

Ongrid is an open-source ops AI agent from the ongridio organization that reasons over metrics, logs, traces, topology, source code and a knowledge base to investigate infrastructure problems. Operators interact with it from chat channels such as Slack or Telegram, and it returns an evidence-backed answer rather than a transcript. The latest release listed is v0.18.0, and the repository is licensed under AGPL-3.0.

## What It Is

Ongrid is an AIOps and incident-response agent for operations teams. It connects to observability data (Prometheus, Loki, Tempo, Grafana, OpenTelemetry) and uses tools to query that data, walk a topology map, and correlate signals to a root cause, which it can pin to a source-code line. A coordinator agent dispatches work to specialist sub-agents (SRE, network, database), and an alert can trigger an investigator that spawns an RCA worker and writes the cause back to chat.

## How It Works

Edge components on each host dial out through a single outbound tunnel, so no inbound ports are required on hosts. The agent uses read-only host tools (a bash sandbox plus 26+ inspection tools) with every call audited, and a browser SSH feature provides a reverse-tunnel shell. Risky actions can be placed behind approvals through a write gate. A workflow builder wires triggers, agents, tools, conditions and notifications into repeatable automations, and external MCP servers can be registered so their tools are available to agents and workflows.

## Models, Channels and Knowledge

Users bring their own model: Anthropic, OpenAI, GLM, DeepSeek, Gemini, Kimi or any OpenAI-compatible relay. Two-way channels include Slack, Telegram, Larksuite, DingTalk, WeCom and webhooks, with per-channel locale. A knowledge vault indexes runbooks, notes, incident history and repositories for RAG search. Kubernetes lifecycle management and network device management (SNMP-verified discovery and interface polling) are also documented.

## Setup Path

Download the release archive for linux-amd64 or linux-arm64, extract it and run install.sh on Ubuntu 22.04+, Debian 12+ or RHEL/Rocky 9. The install brings up the full stack: manager, MySQL, Prometheus, Loki, Tempo, Grafana and Qdrant. It can later be pointed at existing Prometheus, Loki or Tempo instances.

## Features
- Coordinator and specialist agents (SRE, network, DB)
- Auto-investigation on alert
- Root-cause analysis across metrics, logs, traces, topology and source code
- Zero inbound ports via outbound edge tunnel
- Browser SSH with audit
- Self-host install with full observability stack
- Bring your own model with hot routing
- Two-way chat channels with per-channel locale
- Read-only host tools with audited calls
- Kubernetes lifecycle management
- Network device management via SNMP
- Workflow builder
- MCP server registration
- Knowledge vault with RAG search
- Approval and write gate for risky actions

## Integrations
Prometheus, Grafana, Loki, Tempo, OpenTelemetry, Qdrant, Slack, Telegram, Larksuite, DingTalk, WeCom, Webhook, Anthropic, OpenAI, Gemini, DeepSeek, Zhipu GLM, Kimi, Kubernetes, MCP

## Platforms
LINUX, WEB, API, CLI

## Pricing
Open Source

## Version
v0.18.0

## Links
- Website: https://ongrid.cloud
- Documentation: https://ongrid.cloud/docs/get-started/introduction
- Repository: https://github.com/ongridio/ongrid
- EveryDev.ai: https://www.everydev.ai/tools/ongrid