fable-jev
Pairs Jev with the get-fable lifecycle harness to give coding agents fast tool and skill routing and verbatim-preserving compaction.
Overview
Fable-Jev pairs TypeSafe's Jev with get-fable, a deterministic lifecycle engine for autonomous coding agents. It argues that agents waste seconds and thousands of tokens using a large reasoning model to choose the next tool or skill, that keyword routers break on ambiguous instructions, and that LLM summarisation during compaction loses file paths, flags and error codes.
In its dual-system design, Jev provides fast System 1 reflexes such as routing and keep-or-drop decisions, while get-fable supplies deterministic System 2 safety gates. Its stated golden invariant is that Jev is allowed to be wrong but the system is never allowed to become unsafe as a result. The repository links architecture, fast compaction, playbook, benchmark and roadmap documents and an awesome list. The README is written in a promotional style, and claims are best checked against the linked benchmarks.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which tool or skill should the agent trigger next?(inferred) | — |
| Yes / no | Should this context item be kept during compaction?(inferred) | yes · no |
How it works
- Agent reaches a routing or compaction point
- Jev makes a fast typed decision
- get-fable's deterministic gates validate the outcome
- Agent proceeds; unsafe outcomes are blocked by the gates
Strengths
- Fast routing without long chain-of-thought
- Deterministic safety gates around model decisions
- Verbatim-preserving compaction approach
Caveats
- Promotional README; verify benchmark claims
- Depends on the get-fable harness
Getting started
See docs/PLAYBOOK.md and the parent get-fable repository for setup.At a glance
- Type
- Project
- Status
- prototype
- Language
- TypeScript
- Runtime
- agent harness extension
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
⚡ Sub-100ms cognitive reflexes for autonomous coding agents. Powered by TypeSafe AI's Jev & get-fable.
hellogumbo/awesome-jev
Frequently asked questions
What is get-fable?
A deterministic lifecycle engine for coding agents that provides safety gates.
What does Jev do here?
Fast reflex decisions such as tool routing and compaction keep-or-drop.
What is the golden invariant?
Jev may be wrong, but the system must never become unsafe because of it.
Related projects
Cognitive control layer for OpenCode that uses Jev to compute GO, DELIBERATE, VERIFY or BLOCK for agent actions, observing by default.
ProjectbetaTypeScriptOpen-source Jev skills for Claude Code, Codex, Cursor and Grok: one client to gate, route, guard, verify, triage, review and act.
ProjectbetaTypeScriptToolkit where Jev condenses large MCP responses for Claude Code and maps Codex requests to reviewed, fixed repository operations.
ProjectbetaTypeScriptClaude Code plugin with three Jev tools: verbatim context compaction, per-edit model routing and per-prompt skill selection.
ProjectbetaPythonPi extension that strips the inlined skill list and asks Jev which single skill, if any, to load per turn.
Agent skillbetaTypeScriptGives an agent one tool that uses Jev to search a nested capability tree and return only the documented commands a task needs.
ProjectprototypeTypeScript