Organization repository on GitHub
A programmable tool and agent runtime for Pi
- Forks
- 36
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Agents & MCP
Pi Fabric
A programmable runtime that gives the Pi coding agent one fabric_exec tool: Pi writes a checked TypeScript program that composes core tools, MCP servers and agents into branches and fan-out, executed in an isolated QuickJS sandbox with approvals, budgets and durable coordination.
- Stars
- 292
- Language
- TypeScript
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- monotykamary
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- TypeScript
- Capabilities
- Open source
Highlights
- Pi writes one TypeScript program per request; it is statically checked, then run in an isolated QuickJS kernel where tools and MCP servers are callable.
- A sandboxed Python (Monty) kernel is selectable via executor.kernel, and CPython is an explicit escape hatch that fails closed without OS isolation.
- One runtime covers agents, durable resident workers, event-driven actors, councils, workflows, mesh topics and bounded recursive queries.
- A Fabric-powered agent scored 100% across all 25 ARC-AGI-3 environments in one 22.4-hour session with $1,349 in model spend.
- Guardrails cover approvals, isolation, timeouts, concurrency, recursion depth and shared cost budgets, with a native TUI dashboard.
Quickstart
json
{ "executor": { "kernel": "python" } }Watch out
Copy reviewed against the repo README and GitHub metadata on 2026-10-15; the ARC-AGI-3 result is the project's own public scorecard claim and the captured README does not show the install command, only the Node.js 24+ and Pi 1.0.0+ requirements.
Questions about this entry
Short answers drawn from the linked project or page.
A programmable tool and agent runtime for Pi. Instead of many tool calls, Pi writes one checked TypeScript program that composes core tools, MCP servers and captured extension tools, then returns just the result to the conversation.
By default, TypeScript runs in an isolated QuickJS kernel. You can select a sandboxed Python (Monty) kernel with executor.kernel: python; Monty has VM resource limits and no ambient filesystem, network or process access, and is not CPython.
Only for trusted native code that needs full Python. It requires executor.pythonRuntime: cpython plus macOS sandbox-exec or Linux bwrap isolation, and it runs with full OS privileges outside schema enforcement.
A Fabric-powered agent won all 25 environments in one 22.4-hour session with 4 minutes of human time and $1,349 in model spend, according to the linked ARC Prize scorecard.
One-shot workers, durable resident agents, persistent event-driven actors, councils and bounded recursive queries, plus phased workflows, durable topics and compare-and-swap state for coordination.
Node.js 24+ and Pi 1.0.0+. Monty's optional native package installs on supported platforms and full CPython requires 3.10+; Fabric warns when the host is older than the required loadout.
The package is MIT licensed, but model usage is billed by whichever Pi provider you use — the ARC-AGI-3 run, for example, reported $1,349 in model spend. Fabric also lets you set shared cost budgets as a guardrail.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Jev in front of the HeyGen MCP
X: Playwright CLI + Jev vs Playwright MCP
X: A local MCP server that cuts computer use from 23s to 0.8s. Cartwright built Cortex, a local MCP server where…
X: A side panel that drives any site's WebMCP tools
More Jev resources like this
An MCP server that gives coding agents only the JS/TS code a task needs: the TypeScript compiler follows calls, types, callers and tests, Jev judges relevance, and results come back within a token budget.
Jev Agents & MCP#community#catalog-import#logicrw
Community
123wwwa/jevtraceTypeScript
123wwwa/jevtraceTypeScriptSix evidence-oriented tools for the pi and omp coding agents, also usable from any MCP client: navigate unfamiliar code, ask typed questions about repository evidence, review completed changes, and select existing tests.
Jev Agents & MCP#community#catalog-import#logicrw
Community
nomenak/jev-toolsTypeScript
nomenak/jev-toolsTypeScriptAn MCP relay that adds use_jev to every configured server: Jev picks the tool calls toward a goal, the calling model writes any text values, and the relay executes the steps and returns a trace.
Jev Agents & MCP#community#catalog-import#logicrw
Community
chy4pro/jev-in-mcpTypeScript
chy4pro/jev-in-mcpTypeScriptRule-based code review for AI coding agents, powered by TypeSafe Jev. Three skills — review a PR, review anything, or loop until the engine reports clean — share one local MCP review engine that never reads your repository; the calling agent supplies the diff and the business context.
Jev Agents & MCP#community#catalog-import#logicrw
Community
thebous/jev-flash-reviewTypeScript
thebous/jev-flash-reviewTypeScriptKody is an MCP-first cloud for agents, built on Cloudflare Workers, that keeps memory, keys, code, and automations portable across MCP hosts with a compact search-and-execute tool surface instead of a large static catalog.
Jev Agents & MCP#community#catalog-import#logicrw
Community
kentcdodds/kodyTypeScript
kentcdodds/kodyTypeScriptA free, open-source Claude Code plugin that adds skills, agents and hooks for review, testing, security and memory: install ork for the stable v9.x line or ork-alpha for v10, which ships daily, then run /ork:setup to scan your repo and recommend skills.
Jev Agents & MCP#community#catalog-import#logicrw
Community
yonatangross/orchestkitTypeScript
yonatangross/orchestkitTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.