ensō#

ensō (円相) — the Zen circle, drawn in one breath. Imperfect by design; the irregularities are the point.

A small, single-binary, terminal-first agentic coding agent in Go (binary: enso).

ensō talks to any OpenAI-compatible chat endpoint and gives it the tools to read and edit your project, run shell commands, search the codebase, fetch URLs, and orchestrate sub-agents. Sessions persist to SQLite and resume across crashes. Opt into [backend] type = "podman" (or "lima") and the whole agent runs inside a per-project container or VM so it can’t escape the project directory.

Why ensō#

• Local-first by design. Built around llama.cpp’s llama-server running Qwen3.6-35B-A3B on a single RTX 3090. Anything that speaks the OpenAI Chat Completions wire format works.
• Single static binary, no CGO. SQLite via modernc.org/sqlite, pure-Go LSP and JSON-RPC. Builds for Linux, macOS, and Windows without toolchain gymnastics.
• Persistent everything. Every user message, assistant reply, tool call, and event is written to SQLite before the UI renders it. kill -9 mid-tool-call and the session resumes with the interrupted call surfaced as a synthetic tool result.
• Container / VM isolation, opt-in. Set [backend] type = "podman" (or "lima") and the whole agent runs in a per-project container/VM with the cwd bind-mounted at its real path. File tools (read/write/edit/grep/glob) refuse paths outside cwd regardless. Off by default ("local").
• Real LSP. Configure any language server under [lsp.<name>] and the agent gets lsp_hover, lsp_definition, lsp_references, and lsp_diagnostics tools.

Where to start#

• Install — get a binary on your machine.
• Quickstart — first run end-to-end in five minutes.
• User guide — one page per feature.
• Config reference — every field in config.toml.
• Architecture — what’s inside the binary.

Project status#

v2 is shipped — the TUI was migrated from tview onto Bubble Tea v2

• Lipgloss + Glamour for scrollback-native rendering. The binary is in daily use; per-release notes live in CHANGELOG.md, soak-test risks are documented in the architecture page, and the non-goals — what’s intentionally not built and why — live in AGENTS.md at the repo root.