Power your agent
คำสั่งของผู้เผยแพร่เอง ไม่ใช่เอนด์พอยต์ของ PlugMCP
Host a copy with our keys and logs.
npx -y [email protected]Claude Code ไม่มีลิงก์ติดตั้งสำหรับคำสั่งแบบโลคัล ให้คัดลอกคำสั่งด้านล่างแล้วรันในเทอร์มินัล
claude mcp add jeopi-mnemopi -- npx -y jeopi-mnemopi@16.6.25รันในเทอร์มินัล
{ "mcpServers": { "jeopi-mnemopi": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}claude_desktop_config.json
URL แบบรีโมตหรือคำสั่งบนเครื่องของคุณ? การติดตั้งสองแบบต่างกันอย่างไร
รันบน PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
เปิดจากผู้เผยแพร่
Local SQLite memory engine for Oh My Pi agents
| starts working |
| crystallizes goal / constraints / checkable acceptance criteria first — "make it better" gets sharpened, not accepted |
| Planning | plan agent | plan agent + blocking critic gate: okay / iterate / reject verdict, schema-enforced — and runtime-enforced: after a non-okay verdict the task tool refuses to spawn execution agents and write/edit lock the working tree until a fresh okay (or a new user message) |
| Review | reviewer (patch bugs) | reviewer + architect: severity-rated structural verdict that is invalid without an inspected[] evidence list |
| Failure | retry | failure-lesson loop — capture what the failure proved, change the next attempt, split stuck subgoals; no apology loops. Runtime 3-strike counter closes the critic-iterate loop before it can spin |
| "Done" | tests pass | artifact gate + independent grader — a criterion with no command + observed result is reported unresolved, never implied met. goal({op:"complete"}) runs two checks in sequence, both runtime-enforced, neither skippable by the same agent that did the work: (1) a deterministic evidence gate — rejected unless a test/build/typecheck/lint run succeeded after the last file mutation; (2) a goal-verifier sub-agent — fresh context, no exposure to the maker's transcript/thinking, re-inspects the diff and re-runs the claimed verification independently, verdicts okay/iterate/reject. Either gate bounces once with a corrective message before an unchanged retry passes (docs-only / verifier-false-positive escape hatch); any new mutation or verification evidence re-arms both |
| Pipeline | ad-hoc | /jeo: interview → frozen seed → plan → critic gate → bounded execution → artifact-gated verification |
| Identity | omp | jeopi binary; config directory renamed ~/.omp → ~/.jeopi (jeopi config migrate-legacy moves existing auth/sessions/settings over) |
Real gates, no theater. The rest of this README is the engine both share.
Bun (recommended)
bun install -g jeopi-cli
jeopi --version # jeopi/16.x
npm
npm install -g jeopi-cli
From source
git clone https://github.com/akillness/jeopi.git && cd jeopi
bun run setup # installs deps, builds natives, links the global `jeopi` command
macOS · Linux · Windows · bun ≥ 1.3.14. Upgrading from omp? Run jeopi config migrate-legacy once to move ~/.omp to ~/.jeopi — then /login once, keep it forever. jeopi is published as unscoped npm packages — CLI: jeopi-cli (binary command: jeopi), libraries: jeopi-* — fully independent of the @oh-my-pi scope, while legacy @oh-my-pi/pi-* plugin imports keep resolving through the built-in compat shim.
The jeo-code pipeline, rebuilt on jeopi's native subagents. One command drives it:
/jeo <what you want built>
interview Socratic ambiguity gate — goal, constraints, out-of-scope,
│ checkable acceptance criteria. Vague criteria are refused.
▼
frozen seed local://jeo-seed.md — immutable; scope changes reopen the
│ interview, never drift silently.
▼
plan read-only `plan` agent; concrete files, sequencing,
│ per-criterion verification.
▼
critic gate read-only `critic` agent; schema-enforced verdict
│ okay / iterate / reject. No okay → no execution. Ever.
│ Hard gate: the RUNTIME blocks execution spawns and
│ working-tree writes until the verdict is okay.
▼
execute bounded `task` subagents; a failed task feeds the lesson
│ into the next attempt instead of retrying unchanged.
▼
verify suite runs once as a global signal; each criterion cites
its command + observed result, or is reported unresolved.
Upstream jeo-code enforces its critic verdict with a state file and hash check between CLI processes. jeopi runs the whole pipeline inside one agent session — so the gate lives in the session runtime instead:
critic subagent's schema-validated verdict is recorded by the session the moment the run completes.iterate or reject, the task tool refuses to spawn any non-read-only agent, and write/edit/patch reject every working-tree mutation (the local:// sandbox stays writable for seeds, plans, and notes).okay for the revised plan, or you send a new message — the user regaining control is the only override.The /jeo pipeline borrows the deep-research playbook for token efficiency and loop stability:
local:// files; assignments and critic submissions pass paths, never re-inlined bodies.The same discipline is welded into the standing agents:
critic — read-only actionability gate. "If you catch yourself softening a real, blocking gap into iterate just to avoid blocking, that softening is the signal the gap is real."architect — severity-rated structural review whose verdict is invalid without the list of files it actually inspected. A clean verdict is not the absence of inspection.task — smallest correct change, subgoal-by-subgoal, verification evidence before done, debug leftovers removed.jeo-skills (146 skills — deep-research, god-tibo-imagen, perfectpixel, ooo, …) installs into jeopi with zero extra linking: jeopi natively discovers ~/.agents/skills/ plus the .claude / .codex / .config/opencode skill dirs, at both user and project scope.
# global install — jeopi picks these up automatically
npx skills add -g https://github.com/akillness/jeo-skills --skill deep-research --skill god-tibo-imagen
# jeopi-only pin (native roots)
# global : ~/.jeopi/agent/skills/<skill>/SKILL.md
# project: .jeopi/skills/<skill>/SKILL.md
Invoke any discovered skill with /skill:<name> or let the agent route to it by description.
jeopi generates its own completion scripts for bash, zsh, and fish from the live command/flag metadata, so they never drift from the actual CLI. Subcommands, flags, and enum values complete statically; model names (--model, --smol, --slow, --plan) resolve against the bundled model catalog and --resume against your on-disk sessions.
# zsh — add to ~/.zshrc (or write the output into a file on your $fpath)
eval "$(jeopi completions zsh)"
# bash — add to ~/.bashrc
eval "$(jeopi completions bash)"
# fish
jeopi completions fish > ~/.config/fish/completions/jeopi.fish
Edits that land on the first attempt. Reads that summarize files instead of dumping their content. Searches that return instantly. Pick any model — jeopi will get it right.
read : summarized snippets · ideal defaults · selector hit ratesearch : fastest in the westlsp : everything your IDE knows, the agent knowsprompts : adjusted relentlessly for each modelOriginally built on Mario Zechner's wonderful Pi, jeopi adds everything you're missing.
Most harnesses give the agent a Python sandbox and call it done. Ours runs persistent Python and a Bun worker, and either kernel can call back into the agent's own tools — read, search, task — over a loopback bridge. The agent loads a CSV with tool.read from inside Python, charts it from JavaScript, and never leaves the cell.
![jeopi TUI: a single eval session with [1/2] pandas describe (Python) printing a real DataFrame.describe() table, followed by [2/2] top scorer (JavaScript) running a reduce. Footer: 'Both kernels ran in one session.'](https://omp.sh/captures/eval.webp)
Ask for a rename and you get a rename. The call goes through workspace/willRenameFiles, so re-exports, barrel files, and aliased imports update before the file moves. Everything your IDE knows, the agent knows.

A C binary segfaults: the agent attaches lldb, steps to the bad pointer, reads the frame. A Go service hangs: it attaches dlv and walks the goroutines. A Python process is wedged: debugpy, pause, inspect, evaluate. Most agents are still sprinkling print statements.

Your rules sit dormant until the model goes off-script. A regex match aborts the stream mid-token, injects the rule as a system reminder, and retries from the same point. You get course-correction without paying context tax on every turn. Injections survive compaction, so the fix sticks.

Split a job across workers and get typed results back. task fans out into isolated worktrees, each worker runs its own tool surface, and the final yield is a schema-validated object the parent reads directly. No prose to parse, no merge conflicts between siblings, no orphaned edits.

Pair a reviewer model to the 'advisor' role and it reads every turn the main agent takes, injecting notes inline — a quiet aside, a concern, or a hard blocker. It runs on its own context and its own model, so it catches what the doer rushed past. The main agent sees the note and course-corrects, or tells you why it won't.

/collab puts your live session on a relay and hands back a link — and a QR. A teammate joins from another terminal with jeopi join, or just opens it in a browser. Share read-write to pair on the same agent, or /collab view for a read-only link anyone can watch but no one can steer. Frames are sealed client-side; the relay never sees your keys.

web_search chains eighteen ranked providers and hands whatever URLs it finds straight to read. Arxiv PDFs, GitHub pages, Stack Overflow threads come back as structured markdown with anchors intact — the same tool surface you use on local files. Cite, follow, quote, never lose where you came from.

Other agents shell out to rg, grep, find, and bash. On many machines those binaries don't exist, and on the ones where they do, every call costs a fork-exec round-trip. jeopi links the real implementations into the process. ripgrep, glob, find: in-process. brush is the bash, with sessions that survive across calls. The same jeopi binary runs on macOS, Linux, and Windows — no WSL bridge.
Get a clear verdict on whether the change ships, with every issue ranked P0 through P3 and scored for confidence. /review spawns dedicated reviewer subagents that sweep branches, single commits, or uncommitted work in parallel. You tackle what blocks release first; nothing important hides in a wall of prose.
Perfect edits, fewer tokens. The model points at anchors instead of retyping the lines it wants to change, so whitespace battles and string-not-found loops just stop happening. Edit a stale file and the anchors diverge — we reject the patch before it corrupts anything. Grok 4 Fast spends 61% fewer output tokens on the same work.
Other harnesses bolt on gh_issue_view, gh_pr_view, gh_search — each with its own parameters the agent has to learn and you have to debug. We skipped that. read already handles paths; PRs are paths. One interface to teach the model, one surface to keep correct.
The agent remembers your codebase between sessions. It writes facts mid-run with retain, pulls them back with recall, and compresses each session into a mental model that loads on the first turn of the next one. Project-scoped by default, so what it learns about this repo stays with this repo.
Run jeopi inside Zed and you get the same agent you drive from the terminal — reading the buffer you're actually looking at, writing through the editor's save path, spawning shells in the editor's terminal. Destructive tools pause for a permission prompt you can answer once and forget. No bridge, no plugin, no second brain to keep in sync.
Every other agent ships an importer and expects you to convert. jeopi reads the eight formats already on disk in their native shape — Cursor MDC, Cline .clinerules, Codex AGENTS.md, Copilot applyTo, and the rest. No migration script, no YAML-to-TOML port, no "supported subset" footnotes. The config your team wrote last quarter still works tonight.
jeopi reads the working tree through git_overview, git_file_diff, and git_hunk, then splits unrelated changes into atomic commits ordered by their dependencies. Cycles are rejected before anything is written. Source files score above tests, docs, and configs, so the headline commit is the one that matters. Lock files are excluded from analysis entirely.
Twelve internal schemes — pr://, issue://, agent://, skill://, rule://, and the rest — resolve transparently inside every FS-shaped tool the agent already calls. read pr://1428 returns the same shape as read src/foo.ts. search walks a diff like a directory. agent://<id>/findings.0.path pulls a field out of a subagent's output by path.
![jeopi TUI reading pr://can1357/oh-my-pi/1063 and then /diff/1, showing hunk headers, added lines, and a [MODIFIED] (+12 -0) summary.](https://omp.sh/captures/pr.webp)
Each merge conflict becomes one URL. The agent writes @theirs, @ours, or @base to conflict://N and the file resolves cleanly. Bulk form: conflict://*.

ast_edit returns a (proposed) card with the replacement count. The change is staged. The agent calls resolve with a reason; the TUI turns it into an Accept card and the disk move happens — atomic, all or nothing.

Stealth's on by default, so pages see a normal user instead of a headless bot. The same API drives any Electron app in place — point it at Slack and the agent reads your DMs the way it reads the web.

32 tools live in the same namespace as read and bash. Pin the active set with --tools read,edit,bash,… and the rest stay hidden but indexed — search_tool_bm25 pulls them back in mid-session when tools.discoveryMode says so.
Files & search
read — files, dirs, archives, SQLite, PDFs, notebooks, URLs, and internal :// schemes through one path.write — create or overwrite a file, archive entry, or SQLite row.edit — hashline patches with content-hash anchors and stale-anchor recovery.ast_edit — structural rewrites previewed before apply, via ast-grep.ast_grep — structural code queries over 50+ tree-sitter grammars.search — regex over files, globs, and internal URLs.find — glob-based path lookup; reach for search when you need content matches.Runtime
bash — workspace shell, with optional PTY or background-job dispatch.eval — persistent Python and JavaScript cells with shared prelude and tool re-entry.ssh — one remote command against a configured host.Code intelligence
lsp — diagnostics, navigation, symbols, renames, code actions, raw requests.debug — drive a DAP session — breakpoints, stepping, threads, stack, variables.Coordination
task — fan out subagents in parallel, optionally workspace-isolated.irc — short prose between live agents in this process.todo — ordered mutations over the session todo list with phase tracking.job — wait on or cancel background jobs.ask — structured follow-up questions for interactive runs.Outside the box
browser — Puppeteer tabs over headless Chromium or CDP-attached apps.web_search — one query across configured providers, returning answer plus citations.github — GitHub CLI ops — repo, PR, issues, code search, Actions run-watch.generate_image — generate or edit raster images via Gemini, GPT, or xAI Grok image models.inspect_image — vision-model analysis of a local image file.tts — text-to-speech via xAI Grok Voice — five built-in voices, WAV or MP3.Memory & state
checkpoint — mark conversation state for a later collapse-and-report.rewind — prune exploratory context, keep a concise report.retain — queue durable facts into the active Hindsight bank.recall — search the Hindsight bank for raw memories.reflect — ask Hindsight to synthesize an answer over the bank.Misc
resolve — apply or discard a queued preview action.search_tool_bm25 — BM25 over the hidden tool index; activates top matches mid-session.Setting-gated, off by default: github, inspect_image, tts, checkpoint, rewind, search_tool_bm25, retain, recall, reflect. Flip them on once, scoped per project.
Roles route work by intent. default for normal turns. smol for cheap subagent fan-out. slow for deep reasoning. plan for plan mode. commit for changelogs. Override at launch with --smol, --slow, or --plan; cycle through the configured models for the active role with Ctrl+P. Swap the active model mid-session with the /model slash command.
Auth tags below: oauth signs in with your provider account, plan routes through a coding-plan subscription, local runs against a local server with the key optional.
Direct APIs and gateways. Mix providers per role.
Anthropic oauth · OpenAI · OpenAI Codex oauth · Google Gemini · Google Antigravity oauth · xAI · Mistral · Groq · Cerebras · Fireworks · Together · Hugging Face · NVIDIA · OpenRouter · Synthetic · Vercel AI Gateway · Cloudflare AI Gateway · Wafer Serverless · Perplexity oauth
Subscription-routed. /login attaches the session.
Cursor oauth · GitHub Copilot oauth · GitLab Duo · Kimi Code plan · Moonshot · MiniMax Coding Plan plan · MiniMax Coding Plan CN plan · Alibaba Coding Plan plan · Qwen Portal · Z.AI / GLM Coding Plan plan · Xiaomi MiMo · Qianfan · NanoGPT · Venice · Kilo · ZenMux · Tencent Cloud MaaS · OpenCode Go · OpenCode Zen
OpenAI-compatible /v1/models, with local runtimes treated as live snapshots instead of stale catalog guesses. jeopi models <provider> and /model refresh LM Studio/Ollama/llama.cpp from the running server, replace removed entries, and keep local instances keyless by default.
Ollama local · Ollama Cloud · LM Studio local · llama.cpp local · vLLM local · LiteLLM
openai-completions, openai-responses, openai-codex-responses, azure-openai-responses, anthropic-messages, google-generative-ai, or google-vertex in ~/.jeopi/agent/models.yml.retry.fallbackChains. When the primary throws 429s or hits a quota wall, the next entry takes the rest of the turn — restored on cooldown.enabledModels and disabledProviders entries to a path: prefix to pin a different model set on one repo without touching the global config. Scoped entries cover the path and everything under it.Full provider & routing reference at docs/providers.md.
web_search is built in, not bolted on. auto walks an eighteen-provider chain; pin one by name if you already pay for it. Behind every hit, site-aware extraction turns GitHub, registries, arXiv, Stack Overflow, and docs into structured markdown — anchors and link targets survive.
Eighteen backends. Pin one, or let auto walk the chain in order.
The agent gets structured content, not stripped HTML.
Pages convert to markdown with link structure intact. The agent can cite, follow, and quote without losing anchors.
Vuln lookups answer with vendor data, not blog summaries.
Four crates, one platform-tagged N-API addon. Search, shell, AST, highlight, PTY, image decode, BPE counting — all in-process on the libuv pool. No fork/exec on the hot path.
pi-natives, pi-shell, pi-ast, pi-isolinux-x64, linux-arm64, darwin-x64, darwin-arm64, win32-x64The table below is a per-module breakdown that intentionally omits glue and tests.
Same engine, five wrappers. jeopi runs the TUI. jeopi -p answers a single prompt and exits. The Node SDK embeds the session in your process. jeopi --mode rpc and jeopi acp hand the wheel to another program over stdio. jeopi messenger connects Discord, Telegram, and Slack bots to headless sessions.
The TUI is the default surface. Tool calls render as cards, edits preview before they land, and ambiguity routes through the ask tool — a structured option picker the agent can call mid-turn. The keyboard handles the rest.
The same prompt cards surface over ACP, so editors get the picker without writing one.

jeopi
Node and TypeScript hosts pull the engine in directly. The package exposes ModelRegistry, SessionManager, createAgentSession, and discoverAuthStorage; the session emits typed events you subscribe to.
import {
ModelRegistry,
SessionManager,
createAgentSession,
discoverAuthStorage,
} from "jeopi-cli";
const auth = await discoverAuthStorage();
const models = new ModelRegistry(auth);
await models.refresh();
const { session } = await createAgentSession({
sessionManager: SessionManager.inMemory(),
authStorage: auth,
modelRegistry: models,
});
await session.prompt("list .ts files");
jeopi --mode rpc
For non-Node embedders, or when you want process isolation. NDJSON commands in, response and event frames out. --mode rpc-ui adds tool cards, selectors, and dialogs as extension_ui_request frames the host must answer.
$ jeopi --mode rpc --no-session
> {"id":"r1","type":"prompt","message":"list .ts files"}
< {"id":"r1","type":"response", ...}
> {"id":"r2","type":"set_model","provider":"anthropic","modelId":"sonnet-4.5"}
> {"id":"r3","type":"abort"}
jeopi acp
The Agent Client Protocol over JSON-RPC. When the editor advertises capabilities, tool I/O routes through it and writes are gated by session/request_permission.
Full reference: docs/sdk.md.
jeopi messenger check|run --config <file>
Connect authorized Discord, Telegram, and Slack bots directly to headless agent sessions for bidirectional text conversations. Discord Gateway and Slack Socket Mode connect over outbound WebSockets; Telegram connects via outbound HTTPS long polling. No public IP, webhook endpoints, cloud relays, or tunnel tools (like ngrok) are required.
{
"cwd": "./workspace",
"sessionDir": "./sessions",
"toolNames": [],
"platforms": {
"telegram": {
"tokenEnv": "JEOPI_TELEGRAM_BOT_TOKEN",
"allowedUserIds": ["TELEGRAM_USER_ID"],
"allowedChannelIds": ["TELEGRAM_CHAT_ID"]
},
"discord": {
"tokenEnv": "JEOPI_DISCORD_BOT_TOKEN",
"allowedUserIds": ["DISCORD_USER_ID"],
"allowedChannelIds": ["DISCORD_CHANNEL_OR_THREAD_ID"]
},
"slack": {
"tokenEnv": "JEOPI_SLACK_BOT_TOKEN",
"appTokenEnv": "JEOPI_SLACK_APP_TOKEN",
"allowedUserIds": ["SLACK_MEMBER_ID"],
"allowedChannelIds": ["SLACK_CHANNEL_OR_DM_ID"]
}
}
}
jeopi messenger check --config /path/to/messenger.json
jeopi messenger run --config /path/to/messenger.json
message_thread_id).xapp-...) and Bot Token (xoxb-...). Automatically threads channel replies.allowedUserIds). Unauthorized users are ignored silently.[platform, accountId, channelId, threadId, senderId] and cwd.toolNames: [] (pure text chat). Tools (read, grep, glob, bash, edit, write) are strict opt-in.Detailed documentation:
Install the jeopi-messenger skill so AI coding agents (jeopi, Claude Code, Codex, Cursor, Aside) can autonomously configure, validate, run, and troubleshoot messenger bots:
# Universal install via skills CLI
npx skills add akillness/jeopi --skill jeopi-messenger
# Or install manually into agent project skills
mkdir -p .agents/skills/jeopi-messenger
cp -r skills/jeopi-messenger/* .agents/skills/jeopi-messenger/
jeopi's lineage: Pi by Mario Zechner → rewritten as the coding-first surface oh-my-pi by Can Bölük → fused with jeo-code's spec-first discipline as jeopi. Sessions, subagents, slash commands, extensions — all TypeScript, all MIT. Shape it from config, hook it from outside, or read the source when you need to.
An extension is a TypeScript module. Same tool API, same slash-command registry, same hotkey table, same TUI primitives the built-ins use. Nothing is reserved.
On first run jeopi inherits whatever is already on disk: rules, skills, and MCP servers from .claude, .cursor, .windsurf, .gemini, .codex, .cline, .github/copilot, and .vscode. No migration script.
Ask jeopi to write the piece you're missing, then /reload-plugins. Keep it local, ship it in a marketplace, or publish it to npm.
jeopi = oh-my-pi's engine × jeo-code's discipline.
From oh-my-pi (and Pi before it):
From jeo-code:
Fresh clones need both workspace dependencies and the local Rust/N-API addon before the source CLI can start.
bun setup
bun dev
bun setup installs Bun workspaces and builds jeopi-natives. Re-run bun run build:native after changing Rust crates or packages/natives.
For a non-interactive smoke check:
bun dev -- --version
/debug opens tools for debugging, reporting, and profiling.
For architecture and contribution guidelines, see packages/coding-agent/DEVELOPMENT.md.
Latest 5 released entries:
google-antigravity discovery cache keys during synchronous ModelRegistry startup and static reloads, and include google-antigravity and openai-codex in authoritative…<system-conventions> prompt tags with standard markdown headers.See packages/coding-agent/CHANGELOG.md for the full history.
Issues and PRs are open at akillness/jeopi. Keep diffs small, name the acceptance criterion your change serves, and bring the verification evidence — the same contract the agent itself is held to. Upstream engine work belongs in can1357/oh-my-pi; jeopi tracks it.
MIT. See LICENSE.
© 2025 Mario Zechner
© 2025-2026 Can Bölük
© 2026 jeopi contributors
Encode intention. Decode software.
{ "mcpServers": { "jeopi-mnemopi": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}เครื่องมือ
ยังไม่มีรายการเครื่องมือ เซิร์ฟเวอร์แบบรีโมตจะได้รายการหลังการตรวจสอบสถานะครั้งแรก
FAQ
คัดลอกคอนฟิก stdio สำหรับ Cursor, Claude Desktop, Claude Code หรือ VS Code ได้ในหน้านี้ คอนฟิกนี้ใช้การติดตั้งของผู้เผยแพร่ ไม่ใช่พร็อกซีของ PlugMCP
ระเบียนนี้ไม่ได้ระบุตัวแปรสภาพแวดล้อมที่จำเป็น
Jeopi Mnemopi ทำงานแบบโลคัลผ่าน stdio
ตรวจสอบล่าสุดเมื่อ 2026-11-16 Local SQLite memory engine for Oh My Pi agents