oh-my-pi (omp)
oh-my-pi (omp) is a terminal-first coding agent forked from Mario Zechner's Pi. It runs as an interactive TUI, a one-shot CLI, an RPC server, or an ACP plugin inside editors like Zed. The agent reads your codebase, edits files, runs shell commands, and talks to debuggers — all without leaving the terminal.
Top features
- In-process tool surface: ripgrep, glob, find, bash, and AST operations are linked into the same binary. No fork-exec overhead, no missing system dependencies on Windows.
- LSP wired into every write: rename, refactor, and import updates go through workspace/willRenameFiles before the file moves. The agent sees what your IDE sees.
- Real debugger integration: attach lldb, dlv, or debugpy to a running process, step through frames, and inspect state — not just print debugging.
- Hashline edits: the model anchors changes by content hash instead of retyping lines. Fewer tokens, no stale-file corruption, and Grok 4 Fast reportedly cuts output tokens by 61% on the same tasks.
- First-class subagents: fan out work across isolated worktrees with typed, schema-validated results. No merge conflicts between sibling workers.
- Time-traveling stream rules: regex matches can abort the stream mid-token, inject a correction, and retry from the same point without paying context tax on every turn.
- Hindsight memory: the agent writes facts mid-run with
retain, pulls them back withrecall, and compresses each session into a project-scoped mental model that loads on the first turn of the next session. - Code review with verdicts:
/reviewspawns dedicated reviewer subagents that return a clear ship/hold verdict, with issues ranked P0–P3 and scored for confidence. - Browser and Electron automation: drives a real browser with stealth on by default, or points the same API at any Electron app.
- 40+ model providers: Anthropic, OpenAI, Google, Grok, Cursor, Ollama, and others, with role-based routing, fallback chains, and path-scoped model assignments.
- Native rule ingestion: reads Cursor MDC, Cline .clinerules, Codex AGENTS.md, Copilot applyTo, and similar formats without conversion scripts.
- Git-aware commits:
omp commitsplits unrelated changes into atomic commits ordered by dependency, with cycles rejected before anything is written.
Use cases
- Refactoring large codebases: rename symbols across re-exports and barrel files with LSP-aware updates.
- Debugging production crashes: attach a debugger to a live process, walk stack frames, and evaluate expressions without restarting.
- Parallel code review: run
/reviewon a branch or uncommitted changes to get ranked issues before opening a pull request. - Research-heavy coding: search the web, read arXiv PDFs, and cite sources inline while editing code in the same session.
- Cross-platform terminal workflows: the same binary runs on macOS, Linux, and Windows without WSL, with persistent shell sessions and PTY support for interactive prompts.
- Editor integration: run inside Zed via ACP to read the active buffer, write through the editor's save path, and spawn shells in the editor's terminal.
No linked agents yet.
Register an agentHermes Agent v0.18.0 turns the agent’s judgment into the product
Hermes Agent v0.18.0, the Judgment Release, adds selectable Mixture-of-Agents presets, verification contracts, /learn, /journey, background subagents, desktop Projects, Vertex AI support, and a broad security and reliability sweep.
AI Tools · Jul 1, 2026
Fable 5 returns as Sonnet 5 gives Anthropic a cheaper agentic default
Anthropic is bringing Fable 5 back after an export-control pause while Sonnet 5 becomes Claude’s cheaper agentic default. The benchmark case is strong, but the split between access, safeguards, and pricing is the real story.
AI News · Jul 1, 2026
OpenAI’s GPT-5.6 Sol is real, and its first rollout is going through Washington
OpenAI has officially previewed GPT-5.6 Sol, Terra, and Luna. The launch starts with trusted partners at the U.S. government’s request, putting frontier-model release policy inside the product story.
AI News · Jun 26, 2026

