Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

Jot

General-purpose agent loop where Jev chooses each tool call and its arguments and Jot executes it until a reply is ready.

Projectprototype
View on GitHub ↗

Overview

Jot calls itself the first general-purpose System One agent for Jev. Jev chooses tools and arguments; Jot runs the loop: execute the tool, put the real result back into context, and ask Jev again until it can reply. The README's example asks it to add 6 and 7 and multiply by 2: Jev picks two calculate calls with arguments, the calculator does the maths, and Jot replies 26. Jev only decides the calls.

The UI shows every tool call, credentials stay on the server, and chats stay in the browser. The code is split into a chat UI with a tool-call trace, a provider-independent agent loop that imports neither React nor Jev, a TypeSafe client package, and a server that wires the Jev adapter and tools together. Local 'drafts' for replies run on Apple Silicon with uv, and it credits pi, Errand, FrequencyWords and jsRealB, suggesting reply text is assembled rather than generated by Jev. It is independent of TypeSafe, and a list source notes no licence file at the time of writing.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich tool should be called next (or reply now)?—
ChoiceWhat argument values should the tool call use?—

How it works

  1. User sends a message
  2. Jev chooses a tool and arguments
  3. Server executes the tool and returns the result to context
  4. Jev decides again until it chooses to reply
  5. UI shows the full tool-call trace

Strengths

  • Tools do exact work such as arithmetic; Jev only routes
  • Loop is provider-independent
  • Full visibility of tool calls

Caveats

  • No licence file reported
  • Local drafts require Apple Silicon
  • Loopback-only demo app

Getting started

npm install, cp .env.example .env (set TYPESAFE_API_KEY), npm run draft:serve and npm run dev, then open localhost:3000.

At a glance

Type
Project
Status
prototype
Language
TypeScript
Runtime
Node server with web UI
Has code
yes
Has tests
yes

Use cases and integrations

How curated lists describe it

  • The first general-purpose System One agent for Jev.

    hellogumbo/awesome-jev

  • A general-purpose agent loop where Jev picks the next move and Jot runs it. No license file at the time of writing.

    MrJev/awesome-jev

Frequently asked questions

Does Jev do the maths?

No. Jev chooses calculator calls; the tool computes the result.

Is it tied to Jev?

The agent loop is provider-independent; the server wires in the Jev adapter.

What do I need to run it?

Node.js, a TypeSafe key and Apple Silicon with uv for local drafts.

Related projects

  • Jev-AI-Skill

    One skill and MCP server for Claude Code, Codex and Hermes where Jev gates large-model turns, routes models and guards Git steps.

    MCP serverbetaPython
  • jev-harness (apixly-ai)

    Python harness for open-ended goals: AI-authored tools and checks, Jev choosing the next tool and arguments from fresh state.

    ProjectbetaPython
  • JevLoop (parkavenue9639)

    Python agent runtime where Jev Choice decisions pick tools and arguments, and low-confidence picks go to an LLM for review.

    ServicebetaPython
  • laya-thalamus

    Decision middleware for AI agents that uses local Laya to pick tools and judge results, falling back to an LLM when unsure.

    ProjectbetaPython
  • Loki

    Self-improving multi-platform agent harness with an optional Jev companion for typed judgments and per-session model routing.

    ServicebetaTypeScript
  • omo-jev-plugin

    OmO/senpi agent plugin where Jev or Span-01 rates which skills and next tools fit the task and spots loops or completion.

    Agent skillbetaTypeScript