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

dotpals

Local desktop pal that verifies what a coding agent really did, with optional Laya or Jev double-checks of unclear test results.

ProjectbetaMIT
View on GitHub ↗

Overview

dotpals is a small floating desktop companion that watches Claude Code, Codex or any agent and reports in plain words which files changed, which commands ran and whether tests really passed, reading test output rather than trusting the agent. Since version 0.10 it also checks the work: when Claude Code's tests fail it is told why and sent back if it tries to finish, commit or push anyway, and faked passes (a lost assertion, an added skip) are caught. Any AI tool can query it over MCP with tools such as check_my_work and ready_to_merge.

The decision-model part is optional and narrow: an unclear test result can be double-checked by Laya running locally or by TypeSafe's Jev in the cloud, which is off by default. Most of dotpals is deterministic log and output parsing; everything stays on your computer unless you enable cloud Jev.

What the decision model decides

TypeQuestionAnswers
Yes / noDid the tests actually pass, given this unclear test output?(inferred)yes · no

How it works

  1. Watch agent sessions and parse file changes, commands and test output
  2. Mark results passed, failed or unclear from the output itself
  3. Optionally ask Laya or Jev to double-check unclear results
  4. Send the agent back if tests fail or were faked

Strengths

  • Evidence-based view of agent work
  • Local by default; cloud Jev optional
  • MCP tools for other agents

Caveats

  • Decision model plays a small, optional role
  • Cloud Jev requires a TypeSafe key

Getting started

npx dotpals@latest setup.

At a glance

Type
Project
Status
beta
Language
TypeScript
Runtime
desktop app + MCP server
Has code
yes
Has tests
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • A desktop pal that watches Claude Code, Codex or any agent and says in plain words which files changed, which commands ran and what failed. Runs locally.

    MrJev/awesome-jev

Frequently asked questions

Does dotpals use Jev?

Optionally, to double-check unclear test results; Laya can do this locally instead, and both are off by default.

What does it check?

Files changed, commands run, and whether tests really passed based on their output.

Does data leave my machine?

Not unless you enable the cloud Jev option.

Related projects

  • Yardsort

    Desktop app running coding agents in parallel git worktrees, with optional Jev judgments of each changed file against the task.

    Appbeta
  • vibecheck-linear-jev

    Linear-backed work ledger for coding agents where a Jev-compatible model verifies done reports, blockers and plan changes against the task.

    ProjectprototypeJavaScript
  • dsh-completion-supervisor

    DeepSeek Harness plugin checking whether an agent's 'done' claim is true, using deterministic evidence plus one batched Jev assessment.

    PluginbetaTypeScript
  • first-pass

    Claude Code plugin with rules and checks that make agents prove their work; an optional Jev judge triages review findings.

    PluginbetaMarkdown + shell
  • interlinked-cli

    Local guard layer for coding agents that blocks dangerous tool calls by deterministic rules, with optional Jev judgments.

    CLIbetaTypeScript
  • reticle

    Runtime verification framework that drives your real app to check a coding agent's 'done', returning pass, fail or cannot tell.

    ServicebetaTypeScript