Ongrid is an operations AI Agent designed to understand infrastructure, pinpoint fault root causes, and execute remediation. Users can interact with it directly from instant messaging channels such as Slack, Telegram, Lark (Feishu), or DingTalk.
Purpose
Ongrid is built for operations scenarios, helping teams automatically collect topology and device context when triggered by alerts or manual questions, perform root cause analysis (RCA), and provide evidence-backed analysis conclusions and concrete next steps. It also supports Kubernetes lifecycle management and network device management.
Main Capabilities
- Coordinator + expert Agents: the coordinator dispatches tasks to sub-agents such as SRE, network, and database.
- Automatic alert investigation: after receiving an alert, it automatically generates an RCA work process and writes the cause back to the chat channel.
- Root cause RCA: traverses the topology, correlates metrics, logs, and traces, and pinpoints the "why" down to the source code line.
- Zero inbound ports: edge components actively dial out, so no ports 22/80/443 need to be opened on hosts.
- Browser SSH: enter any host through a reverse tunnel, with no keys, no jump server, and full auditing.
- One-command self-hosting: install.sh can bring up the complete technology stack.
- Built-in observability: already integrated with Prometheus + Loki + Tempo + Grafana, with queries written by the Agent.
- Bring your own models: supports Anthropic, OpenAI, GLM, DeepSeek, Gemini, and Kimi, with hot-switchable routing.
- Bidirectional IM channels: Slack, Telegram, Larksuite, DingTalk, and WeCom, with per-channel language settings.
- Read-only host tools: bash sandbox + more than 26 inspection tools, with every invocation audited.
- Kubernetes lifecycle: register clusters, inspect workloads and events, manage upgrades, and map Kubernetes resources to the topology.
- Network device management: discover neighbors from edge hosts, verify via SNMP, poll interfaces, and map links from hosts to network devices.
How to Use
Download the latest version for the corresponding server architecture (linux-amd64 or linux-arm64), extract it, and run the installer. It supports Ubuntu 22.04+, Debian 12+, and RHEL/Rocky 9. If users in mainland China encounter slow GitHub access, they can use the corresponding CDN mirror URL. Complete product documentation is available at ongrid.cloud.
Applicable Scenarios
- Alert-driven automatic root cause analysis and troubleshooting.
- Operations collaboration and remote execution through chat channels.
- Kubernetes cluster lifecycle management and network device management.
- Building repeatable, reviewable automated workflows (Workflow Builder).
- Unified management of the operations knowledge base (Knowledge Vault) and generated artifacts (Artifacts Center).
- Safely executing high-risk operations under approval and write gates (Approval and Write Gate).
Limitations
- The license is AGPLv3; see LICENSE for details. Ongrid brand assets are not covered by the AGPLv3 license; see TRADEMARK.md for details.
- Specific fees, pricing, platform support details, performance data, and similar information are not yet provided.
