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

openclaw-typesafe-ai

Independent OpenClaw plugin exposing one explicitly invoked typesafe_decide tool that sends supplied state and typed questions to Jev.

HookbetaMIT
View on GitHub ↗

Overview

openclaw-typesafe-ai adds a single optional agent tool, typesafe_decide, to OpenClaw. An agent calls it explicitly with text or JSON state and typed questions, and the plugin forwards them to POST https://api.typesafe.ai/v1/systemone with jev-latest as the default model. It deliberately registers no model provider, picker entry, hook or background service, because Jev returns decisions rather than prose.

The README contrasts it with another community package that uses runtime hooks for group triage, tool guardrails, routing, audit and compaction: this one submits only caller-supplied data and uses OpenClaw's capability-owned SecretInput for the key. Compatibility is pinned to OpenClaw 2026.9.4 on Node 24.20.0 on macOS arm64, with a verification report describing test coverage and evidence limits. It is an independent, MIT-licensed integration for teams wanting explicit, auditable decisions.

What the decision model decides

TypeQuestionAnswers
OtherAny Choice, Score or Noul question the agent explicitly passes to typesafe_decide—

How it works

  1. Agent decides to call typesafe_decide.
  2. Plugin sends the supplied state and questions to /v1/systemone.
  3. Typed answers return to the agent.
  4. Agent acts on them; no automatic interception occurs.

Strengths

  • Only deliberately supplied data is sent.
  • Minimal surface: no hooks or background services.
  • Credentials handled through OpenClaw's secret references.

Caveats

  • Compatibility claimed only for OpenClaw 2026.9.4.
  • Relies on the agent choosing to call the tool.

Getting started

Install openclaw-typesafe-ai from npm or ClawHub, configure the TypeSafe key as a SecretInput, and enable the typesafe_decide tool for an agent.

At a glance

Type
Hook
Status
beta
Language
TypeScript
Runtime
OpenClaw plugin, Node
Has code
yes
Has tests
yes
License
MIT

Use cases and integrations

How curated lists describe it

Frequently asked questions

Does it intercept messages or tool calls?

No, it only provides a tool the agent calls explicitly.

How is it different from openclaw-plugin-typesafe-ai?

That package uses runtime hooks for automatic policies; this one exposes a single explicit tool.

Which OpenClaw versions are supported?

Only 2026.9.4 is claimed compatible.

Related projects

  • llm-typesafe

    Simon Willison's LLM plugin that adds Jev as a model for yes/no, choice and score questions from the command line.

    PluginproductionPython
  • adk-go-typesafe

    Go integration bringing TypeSafe Choice, Score and Noul to Google's adk-go agents, with a typed client, routing agent and callbacks.

    ProjectbetaGo
  • feelings

    Ruby gem for probabilistic conditionals like Feelings(msg).like?("urgent"), answered by Jev or other decision models.

    ProjectbetaRuby
  • fuzzy-jev

    Rust CLI and library that asks Jev typed questions via OpenRouter and combines the probabilities with fuzzy-logic rules into decisions.

    CLIbetaRust
  • Gut

    Elixir DSL (Gut.feel) that picks an Elixir value for a subject and question, with a Jev evaluation adapter among LLM options.

    ProjectbetaElixir
  • pi-jev-core

    Minimal Pi extension and TypeScript client registering a jev_evaluate tool for noul, choice and score across five Jev platforms.

    ProjectbetaTypeScript