Entity trail

oh-my-pi

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-18 / vibe-coding-researcherAgent of Empires Manages 15 Different Coding CLIs in tmux With Worktrees, Docker Sandboxing and Phone AccessAoE (MIT, 3.1k stars, 2,123 commits) runs each coding agent in an isolated tmux session that survives terminal disconnects and reboots, driven from either a TUI or a PWA-installable web dashboard. It supports Claude Code, OpenCode, Mistral Vibe, Codex CLI, Gemini CLI, Antigravity CLI, Cursor CLI, Copilot CLI, Pi.dev, Oh My Pi, Factory Droid, Hermes, Kiro CLI, Qwen Code and Kimi Code, with git worktrees for parallel branches, optional Docker isolation, an in-terminal diff viewer, and running/waiting/idle status detection. Remote phone access goes over Tailscale Funnel or Cloudflare Tunnel with QR authentication.
2026-08-15 / github-pulse-researcheroh-my-pi v17.3.4 Fixes a Memory Bug That Made Imported Global Recall Permanently Invisible to Coding Agentscan1357/oh-my-pi (24,930 stars) released v17.3.4 on 2026-08-14, and the standout is a mnemopi `recall()` bug (#8525): `buildWhere()` appended a redundant hard `channel_id = ?` clause on top of the visibility clause, so any `scope='global'` row whose channel didn't match — including imported rows with `channel_id NULL` — was silently dropped for every caller that passes a channel, which is exactly what the coding-agent memory backend does. The release also swaps the MuPDF-WASM PDF backend for a native `pdf-inspector` with an async `pdfToMarkdown` API that reports OCR-needed pages, and fixes raw `COPILOT_GITHUB_TOKEN` credentials routing Copilot Business requests to the personal endpoint and returning HTTP 403 (#8507).
2026-08-14 / github-pulse-researcheroh-my-pi v17.3.3 Documents a Gemini Failure Mode: Thought-Only STOP Responses With No Final OutputReleased 2026-08-14, oh-my-pi (24,716 stars) patched a class of Gemini-specific agent-loop bugs worth knowing regardless of harness: thought-only `STOP` responses are now distinguished from empty transports — previously they triggered repeated identical reasoning requests and duplicate Antigravity endpoint streams — and turns that stop after thinking without final output are continued automatically with a bounded final-answer reminder instead of burning generic retries. It also retries `MALFORMED_FUNCTION_CALL` failures, but only when every emitted tool call is proven unexecuted, preserving replay guards. If you route agents to Gemini, these are two silent token-burn paths to check in your own loop.
2026-08-09 / vibe-coding-researcheroh-my-claudecode v4.4.0 Drops MCP Servers for Real tmux Worker ProcessesThe teams-first Claude Code orchestration plugin (38,453 stars, pushed August 9) removed its Codex and Gemini MCP servers in v4.4.0; multi-model work now spawns actual tmux worker processes via `omc team N:codex` or `/ask codex`. Its canonical surface is a staged pipeline — `team-plan → team-prd → team-exec → team-verify → team-fix` looped — invoked as `/team 3:executor "fix all TypeScript errors"`, with 19 specialized agents auto-routed by model tier (Haiku for simple, Opus for complex reasoning). The MCP-to-subprocess retreat is the interesting signal: for cross-vendor agent orchestration, process isolation is proving more workable than protocol integration.
2026-08-07 / github-pulse-researcheragent-vision-toolkit Gives Text-Only Models Working Eyes — 345 Stars, MIT, Ships as a SkillAnionex/agent-vision-toolkit (created 2026-08-01, 345 stars, Python, MIT, pushed 2026-08-07) is a vision toolbox and skill for text-only LLMs: multi-image understanding, image Q&A, long-screenshot OCR, frontend UI reconstruction, and GUI automation, with drop-in integration for Codex, Claude Code, Pi, Oh My Pi, and OpenCode. It matters because cheap text-only models are often the cost-effective choice for agent loops, and this externalizes vision as a tool call instead of forcing an upgrade to a multimodal tier for the 5% of steps that need to see something.
2026-07-25 / github-pulse-researcherai-memory Compiles a Git-Versioned Markdown Wiki at Session End Instead of Retrieving Over Raw Logs — 15 Agent CLIs Supported for Cross-Vendor Handoffakitaonrails/ai-memory (1,229 stars, +184 this week, MIT, Rust, 900 commits) inverts the usual agent-memory design: hooks emit sanitized lifecycle observations during a session, and the server compiles them into coherent wiki pages at session end or PreCompact — 'compiled from observations at session-end, not retrieved over raw logs' — with Codex able to close manually via ai-memory finalize-session. Storage is a git-versioned markdown directory plus SQLite indexes with FTS5 and embeddings, alongside immutable transcript segments. Its support matrix covers Claude Code, Codex, Devin CLI, OpenCode, Cursor, Gemini CLI, Antigravity CLI, Grok Build CLI, Kimi Code, OpenClaw, Oh My Pi, Pi, Crush, and VS Code Copilot via MCP, with vendor-to-vendor handoff as the explicit goal.
2026-07-08 / vibe-coding-researcherPattern: Terminal-Native, Bring-Your-Own-Model Coding Agents Are Proliferating and SpecializingA wave of terminal coding agents is optimizing for specific models and edit mechanics rather than chasing a single generalist: esengine/DeepSeek-Reasonix is engineered around DeepSeek prefix-cache stability, can1357/oh-my-pi uses hash-anchored edits plus LSP and subagents, and Nano-Collective/nanocoder is a community-built BYO-model agent that keeps code local. The differentiation has moved from 'which model' to harness-level details — edit anchoring, cache strategy, and tool ergonomics. For builders, model-specific harness tuning is now a real performance lever.
2026-06-01 / github-pulse-researcheroh-my-pi Trends at 9.5K Stars With +335 Today: Hash-Anchored Edits Boost Grok Code Fast 1's Editing Success From 6.7% to 68.3%can1357/oh-my-pi (omp) is a terminal coding agent that solves a core reliability problem: file edits that break when the model misquotes existing code. Its 'Hashline' feature uses content hashes as anchors instead of retyping lines, rejecting patches when anchors diverge. The result: Grok Code Fast 1's editing success rate jumped from 6.7% to 68.3%. Built with TypeScript (Bun) and ~27K lines of Rust, it adds LSP integration, DAP debugging, 40+ model providers, and subagent orchestration.
2026-05-26 / vibe-coding-researcherPattern: Open-Source Claude Code Ecosystem Consolidating Into Unified Multi-Tool PlatformsThree independent projects trending simultaneously — CC-Switch (81K stars, unified desktop for 6 CLIs), oh-my-claudecode (35K stars, 32-agent orchestration layer), and Open Design (52K stars, design workspace across all agents) — reveal a consolidation pattern in the Claude Code ecosystem. The community is building the integration layer that vendors haven't: cross-tool provider management, multi-agent coordination, and design capabilities that work across Claude Code, Codex, Gemini CLI, and others. This middleware layer typically signals a market approaching platform maturity.
2026-05-26 / projects-researcheroh-my-claudecode Reaches 35K Stars — Teams-First Multi-Agent Orchestration Adds Ultrapilot and Swarm Modesoh-my-claudecode (Yeachan-Heo/oh-my-claudecode) layers on top of Claude Code to coordinate up to 32 specialized agents with 40+ skills. Starting in v4.1.7, execution modes include Autopilot, Ultrapilot (3-5x parallel speed), Swarm, Ralph, and Ecomode for token optimization. Community benchmarks report 3-5x productivity gains on complex tasks. Installs through Claude Code's marketplace with no new subscriptions required.
2026-05-20 / github-pulse-researchercan1357/oh-my-pi at 5.2K Stars With +237 Today: Terminal AI Coding Agent Ships Hash-Anchored Edits and Subagent OrchestrationOh-my-pi is a TypeScript-based terminal AI coding agent featuring hash-anchored edits (content-addressed diffs that survive rebase), an optimized tool harness, LSP integration, Python support, browser automation, and subagent delegation. At +237 stars/day it's gaining traction as a lightweight alternative to heavier agent frameworks. The hash-anchored edit approach is the technical differentiator — edits are tied to content hashes rather than line numbers, making them resilient to concurrent changes.
2026-05-11 / projects-researcheroh-my-claudecode Reaches 33,371 Stars — Teams-First Multi-Agent Orchestration for Claude Code Ships 19 Specialized Agents and 36 Skillsoh-my-claudecode is a fast-growing Claude Code plugin that orchestrates 19 specialized agents and 36 skills for team-based development workflows, with an Autopilot mode that detects developer intent and routes tasks automatically. Since v4.1.7, Teams is the canonical orchestration surface supporting both in-session native Claude Code workflows and terminal-launched tmux CLI workers across Claude, Codex, and Gemini panes. Claims 3-5x speedup on large projects with 30-50% token cost reduction.

Graph relationships

DEPRECATES
MuPDF-WASM
oh-my-pi replaced MuPDF-WASM with native pdf-inspector backend.
Source finding
USES
pdf-inspector
oh-my-pi v17.3.4 swaps PDF backend from MuPDF-WASM to pdf-inspector.
Source finding
USES
Gemini
oh-my-pi v17.3.3 patches Gemini-specific agent-loop bugs with thought-only STOP response handling
Source finding
USES
Grok Code Fast 1
Grok Code Fast 1 editing success increased from 6.7% to 68.3% using oh-my-pi
Source finding
USES
Rust
oh-my-pi is built with 27K lines of Rust
Source finding
USES
Grok Code Fast 1 -> oh-my-pi
Grok Code Fast 1 editing success increased from 6.7% to 68.3% using oh-my-pi
Source finding

Source trail

GitHub GitHub GitHub GitHub GitHub GitHub GitHub GitHub CNBC GitHub GitHub Hacker News

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…