reticle
Runtime verification framework that drives your real app to check a coding agent's 'done', returning pass, fail or cannot tell.
Overview
Reticle checks whether a coding agent's claim of 'done' is true by driving the real, running app and reading what happened. A small dev-only SDK inside the app gives the agent tools to look, act, observe and assert, and Reticle sees what the screen hides: the request a click sent, its response, app state, console output and each element's source line. Every check ends in a verdict, verified yes, no or unknown, with the evidence and the file:line to fix.
The Jev connection comes from the directory listing, which says Reticle supports Jev as the harness driver for non-generative assertions; the fetched README excerpt does not describe it. In that role, Jev would answer typed assertion questions over observed evidence instead of a generative model writing a judgment. The SDK, adapters, core and engine are Apache-2.0, while the server and CLI are FSL-1.1-ALv2 (Apache-2.0 after two years), and packages run no install scripts.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Given the observed requests, state and console, did the change work: yes, no or cannot tell?(inferred) | yes · no · unknown |
How it works
- Agent finishes a change and claims done
- Reticle drives the running app through its SDK
- Observe requests, responses, state and console
- Assert and return a verdict with evidence and file:line
Strengths
- Catches failures the agent never looked for
- Verdict includes the deciding evidence
- Explicit 'cannot tell' rather than a forced pass
Caveats
- Jev support is described only in the list summary
- Server and CLI are source-available, not fully open source yet
Getting started
npm install -g @reticlehq/server, which puts the reticle command on your PATH.At a glance
- Type
- Service
- Status
- beta
- Language
- TypeScript
- Runtime
- Dev SDK + MCP server/CLI
- Has code
- yes
- License
- Apache-2.0 / FSL-1.1-ALv2
Use cases and integrations
How curated lists describe it
An end-to-end runtime perception and verification framework for coding agents that drives real apps and supports Jev as the harness driver for non-generative assertions.
logicrw/awesome-jev-projectsAI agents can generate code, but still struggle to understand what they build.
wh000wh000/awesome-jev-live
Frequently asked questions
What does Reticle decide?
Whether an agent's claimed fix actually works in the running app: pass, fail or cannot tell.
Where does Jev fit?
According to the directory listing, Jev can drive non-generative assertions in the harness.
How is it different from Playwright?
It reads the app from the inside (requests, state, console, source lines) rather than only the screen.
Related projects
Local desktop pal that verifies what a coding agent really did, with optional Laya or Jev double-checks of unclear test results.
ProjectbetaTypeScriptJev-first CLI coding agent that routes a request to a deterministic workflow which asks Jev fixed-choice questions about bounded evidence.
ProjectprototypeTypeScriptGoal-driven browser agent where each step is one Jev choice over the page's controls, then reviews the run for root causes.
AppbetaTypeScriptFlutter integration-test package where Jev picks the next registered UI action and judges whether the test goal is reached.
LibrarybetaDartPlain-English end-to-end tests for Android and iOS apps where Jev works out the taps and a lockfile replays every decision exactly.
AppbetaPythonBrowser agent with no LLM: code enumerates options from the page and goal, and Jev only picks actions, values and when to stop.
ProjectprototypePython