NewTocket
A command-line toolkit for agent handoffs, workspace context, and developer workflows, with optional Jev decisions for routing and checks.
Added to Jevfast
NewHow it uses Jev
Tocket turns caller-supplied state and named Choice, Noul, or Score questions into one Jev request. It saves the typed answer and handoff decision in the shared project notebook. A separate worker executes the selected move; the default stub and shadow modes are explicitly marked as non-executing.
What Jev decides
Select the next notebook move and review needExample answers · not a recorded Jev response · Source ↗
Question 1 · next
YOUR APP
INSTRUCTION
Pick one value for “next”.
STATE
Repository fixture context: a user requests the next move for a documentation goal. The command supplies next with research/write/review options and needs_human_review in one request. Highlighted answers are illustrative; the fixture also supports a deterministic stub.
JEV · CHOICE
- research
- write
- review
Question 2 · needs human review
YOUR APP
INSTRUCTION
How true is “needs_human_review”? 0 = false, 1 = true.
STATE
Repository fixture context: a user requests the next move for a documentation goal. The command supplies next with research/write/review options and needs_human_review in one request. Highlighted answers are illustrative; the fixture also supports a deterministic stub.
JEV · NOUL
YesNo
App workflow
- Provide state and questions
The caller supplies notebook context and named Choice, Noul, or Score questions to the decide command.
- Ask or run an explicit stub
Live mode sends the typed request to Jev. Dry-run and missing credentials use a marked deterministic stub; shadow mode is log-only.
- Save the handoff
The CLI writes a decision receipt under .context/decisions and applies the documented confidence rule to select a notebook destination.
- Let the worker act
A separate worker reads the saved decision; Tocket stages or applies the notebook receipt instead of executing the worker task.
Why it is interesting
It exposes Jev decisions through a developer CLI and uses fixture-based evaluations to inspect how decisions affect handoffs.