# Blazorly

> Blazorly is an open-source agentic coding harness you run locally. Chat UI or headless CLI. Bring your own model. MIT licensed.

Blazorly is the runtime around a coding model: tools, durable sessions, sandbox presets, and multi-agent fan-out. It is not a hosted cloud agent and not an IDE. Install is a self-contained binary for Linux, macOS, and Windows (x64 and arm64).

## Product
- [Home](https://blazorly.dev/): What Blazorly is, demo, install commands
- [Features](https://blazorly.dev/features/): UI, tools, swarms, durability, System One, CLI
- [Install](https://blazorly.dev/install/): One-command install, settings, providers
- [Compare](https://blazorly.dev/compare/): vs Claude Code, Cursor, OpenCode, Aider
- [About](https://blazorly.dev/about/): License, evidence, source

## Guides
- [What is an agentic coding harness?](https://blazorly.dev/guides/what-is-an-agent-harness/)
- [Getting started](https://blazorly.dev/guides/getting-started/)
- [Settings tabs](https://blazorly.dev/guides/settings/): Model & API, custom providers, sessions, capabilities, sandbox
- [Chat vs Trajectory](https://blazorly.dev/guides/session-views/)
- [Multi-agent orchestration](https://blazorly.dev/guides/multi-agent/)
- [Durable sessions](https://blazorly.dev/guides/durable-sessions/)
- [Permissions and sandbox](https://blazorly.dev/guides/permissions/)
- [Headless, CI, and editors](https://blazorly.dev/guides/headless-ci/)

## Key facts
- Single binary, six platforms; no SDK, no Docker
- Web UI at http://localhost:5080; `blazorly run` for CI (exit 0/2/3)
- Permission presets: full-access, workspace-write, read-only
- Interruption is durable: median 0.3 ms mid-stream cancel; 132,010-event session cold-replays in 1.29 s
- Providers: DeepSeek, OpenAI, Anthropic, xAI, Ollama, LM Studio, OpenAI-compatible routes
- Source: https://github.com/deepakkumar1984/blazorly-harness
- Full text for agents: https://blazorly.dev/llms-full.txt