Three JSON routes made my review board agentic
An agent handoff is a route that answers three questions: which version is current, what feedback is still open on it, and what files produced it. Write those three answers as JSON and most of the plumbing disappears, along with any client library, state machine or ceremony sitting between the model and the data. The build was one evening, October 28, 2026. A video-review workspace shaped after…
Three JSON routes were designed by an agent handoff system, simplifying the process by answering three key questions: which version is current, what feedback is still open, and what files produced it. The entire build process was completed in a single evening on October 28, 2026. The system utilized Flask (flask.do) as its foundation, with a limited dependency list consisting only of the standard library.
The agent's API, located at 127.0.0.1:8791, consisted of just two Python files: one for the API and another for the UI. The agent's interface displayed the message "Reply with Hermes," while the feature list included instructions for interacting with Claude and Opus 5.5. The agent's single-user loopback setup focused on simplicity and minimalism, removing features such as teams, permissions, share links, pricing tiers, and cloud storage.
When an agent processes a request, it receives a job description in the form of a GET request to /api/agent/pid with the latest version, unresolved feedback, and a context bundle. The agent then filters open feedback based on the latest version, ignoring resolved feedback once a new version becomes available. The agent never redoes or questions its own attempts, as the resolve POST response confirms success.
The agent's board displayed a pre-block with a curl command and project goal, allowing users to copy and paste the information directly. The browser's marker system was adapted to work with timecodes, converting percentage values to milliseconds based on the clip's duration. Feedback could be filed in four different formats, each containing specific data such as text, strokes, voice takes, or references.
The system's durability mechanism ensured that all relevant information was stored in a single, atomic write. The agent could then retrieve the necessary context and respond accordingly, maintaining consistency throughout the process. With real-world testing, the agent was able to handle version changes seamlessly, ensuring a smooth user experience.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.