omp
can1357 (Stencil Labs) · ai coding agents · reviewed by Stan K. · Source: data snapshot · data as of 2026-11-10
oh-my-pi — batteries-included Pi fork with a native Rust core, LSP and a debugger wired into the terminal.
#4 of 15 in AI Coding Agents
ToolScore breakdown — why 78.5/100
10 criteria × 10 points. Low pillars are visible on purpose — that's the point of a score you can audit. Methodology →
What it does well
- ▸LSP (14 ops) + DAP debugger (28 ops) + subagents + plan mode built in — "the IDE wired in" is literal
- ▸Native Rust core; sessions saved locally by default, no analytics
- ▸60+ providers incl. subscription OAuth and local models (Ollama/vLLM auto-discover)
- ▸HN: "Currently the very best" / "dead simple to start" (vs OpenCode)
Where it costs you
- ▸10-month-old project moving extremely fast — 4.7k open issues, frequent breaking changes
- ▸No formal privacy/retention policy page (docs/privacy is 404)
- ▸Community support only, small team behind it
Hidden traps — the stuff the pricing page doesn't say
Using Claude Pro/Max, Copilot or (formerly) Antigravity through omp is an unofficial client. Google actually disabled accounts; omp had to disable Antigravity logins (PR #18622).
Opening a repository loads its extensions, hooks and MCP config by design — a cloned repo is a code-execution vector. Mitigate with --no-extensions or mcp.enableProjectConfig: false.
Retention/training questions are answered only by "sessions are saved locally" — a red flag for enterprise compliance.
Know the limits before you pay: read the traps above, then decide with eyes open.
Visit official siteAffiliate link where available — never affects the score.
- official omp official site verified
- official GitHub repo can1357/oh-my-pi (38.6k stars, MIT) verified
- community HN launch thread ("Currently the very best") verified
- official Antigravity ban incident (PR #18622 + HN) verified
Data as of 2026-11-10. Prices and limits change monthly — report anything stale. How we verify: methodology.
Closest alternatives
Google's open-source terminal agent with the most generous free tier in the category and a 1M-token
Open-source minimalist CLI agent — 121k stars, model-agnostic, "the Emacs of coding agents".
The open-source IDE assistant layer — autocomplete, chat, and agents over any model you point it at.
omp review
In-depth analysis · last updated 2026-11-10
Pi with the batteries included
omp (oh-my-pi) is what happens when a Pi fork refuses to be minimalist: subagents, plan mode, hindsight memory, snapcompact context compression — all on a native Rust core, all in the terminal.
The IDE, literally wired in
"IDE wired in" is not marketing here: a full LSP (14 operations — workspace renames, diagnostics) and a DAP debugger (28 operations) run inside the agent, and ACP mode puts the same agent inside Zed. HN's verdict: "Currently the very best"; converts from OpenCode call it "dead simple to start".
Provider matrix without equal
Sixty-plus providers — API keys, subscription OAuth (Claude Pro/Max, ChatGPT/Codex, Copilot, Cursor), cloud gateways, and local Ollama/llama.cpp/LM Studio/vLLM with auto-discovery. The widest BYOK surface we score.
The honest downside
It's ten months old, shipping 37k commits with 4.7k open issues — expect breakage. There is no formal privacy policy page. And riding subscriptions through an unofficial client carries real ban risk: Google disabled accounts until omp pulled Antigravity support.
Verdict
For power users who want a maximal harness at $0 — 79.5/100. Pin your version, use API keys over subscription OAuth where possible, and read a repo's .omp config before opening it.