Siclaw v0.3.4: AI SRE Platform for Infrastructure Diagnostics

What changed

Siclaw has released version v0.3.4 of its open-source AI-powered SRE platform, which focuses on read-only infrastructure diagnostics. This platform is designed to help DevOps and SRE teams gather evidence, form hypotheses, validate them, and ultimately arrive at a clear root-cause analysis without making direct changes to the environment. Users can describe a problem in natural language, and Siclaw will investigate it via its web UI, team chat channels, or a non-interactive command-line interface (CLI).

The platform features a four-phase workflow for deep investigation, encompassing evidence gathering, hypothesis testing, and root-cause analysis. It incorporates an investigation memory that learns from past incidents to improve future diagnostic sessions. A key principle of Siclaw is its read-only approach, ensuring that investigations and recommendations are made without altering the live infrastructure. It also supports team workflows with shared web UI access, credentials, channels, triggers, and scheduled patrols. Furthermore, Siclaw allows for the creation of reusable skills by turning repeated diagnostic playbooks into reviewable runbooks and is extensible through MCP (Model Context Protocol) for connecting external tools and data sources. Access is available through a web UI, chat channels, or the CLI. Agent tracing capabilities allow for exporting agent behavior, including LLM calls, tools used, and tokens consumed, to backends like Langfuse or Phoenix via OTLP, configurable within the web UI with live hot-reloading.

The architecture comprises a control plane (Portal, Gateway, shared DB) for storing curated agents, skills, knowledge, MCP servers, and credentials. Investigation sessions utilize an AgentBox, which runs as a Pod in Kubernetes or in-process during local development. The headless CLI embeds the same core functionality. The Agent Brain executes a Deep Investigation Engine against its bound capabilities, operating in a read-only manner across all targeted systems.

For local usage, Siclaw requires Node.js version 22.19.0 or higher and npm. A quick start involves installing Siclaw globally via npm and running siclaw local to start a lightweight web UI backed by SQLite, suitable for VMs or laptops. This local setup creates a default admin/admin login and allows configuration of models and Kubernetes clusters through the web UI. For non-interactive diagnostic runs, the CLI can be used from the same working directory, invoking agents with specific prompts. For team or enterprise deployments, Siclaw can be deployed to Kubernetes using Helm charts, requiring container images for runtime, portal, and agentbox, and a MySQL database URL for production.

Configuration for the headless CLI involves either a reachable local Portal for providers and agent resources or a .siclaw/config/settings.json file for standalone use. For cluster and host access, credentials must be imported into the local Web UI. Investigation traces are saved to .siclaw/traces/. The local server and Kubernetes deployments centralize configuration through the web UI, covering LLM providers, kubeconfigs, SSH hosts, chat channels (Slack, Lark, Discord, Telegram), MCP servers, user management, and scheduled tasks.

Why it matters for builders

This release of Siclaw v0.3.4 offers builders a powerful, open-source AI agent specifically tailored for SRE and DevOps tasks. Its read-only diagnostic approach minimizes risk during investigations, allowing teams to confidently explore issues without fear of unintended system changes. The platform's extensibility through MCP and support for agent tracing provides deep integration capabilities with existing observability and LLM tooling, enabling custom diagnostic workflows and enhanced debugging processes.

Practical impact

DevOps and SRE teams can immediately begin using Siclaw for incident investigation by setting up the local server or deploying to Kubernetes. Builders can explore the live preview at siclaw.ai/demo to understand its capabilities. The platform encourages the development of reusable diagnostic skills and runbooks, which can be shared within teams to standardize incident response. Developers can also integrate Siclaw's agent tracing with tools like Langfuse for detailed analysis of LLM interactions and tool usage during investigations, optimizing agent performance and debugging complex AI-driven processes.

Caveats and source limits

The provided source information details the features, architecture, and installation of Siclaw v0.3.4. However, specific benchmark results comparing its diagnostic performance against other tools are not included. Pricing information for any potential enterprise features or support is also absent, as is a definitive release date beyond the mention of a "fresh release" and version number. The source primarily consists of GitHub repository content, which serves as an official statement from the developers.