SpecsRelay

SpecsRelay turns clarified AI conversations into reviewed, structured tasks for local coding agents. CAPTURE COMPLETE CONVERSATIONS Open SpecsRelay on a supported AI chat page. Capture starts only after you click the capture button. SpecsRelay preserves user and assistant turns, code blocks, and image references. ORGANIZE AND CLARIFY REQUIREMENTS Use an API provider and model that you configure to convert the conversation into a validated, human-readable coding requirement. SpecsRelay identifies missing decisions, generates follow-up questions, autosaves working drafts, and reveals the next step only when the current one is complete. REVIEW BEFORE HANDOFF Review the task summary, open questions, selected desktop coding agent, and exact plan-only prompt. SpecsRelay requires explicit confirmation before opening or copying the task for the selected local coding agent. WHAT'S NEW IN 1.0.2 • View the captured text and inputs used to organize the current requirement or a saved handoff version. • Review clarification answers, revision instructions, and the previous requirement alongside the saved sources. • See when captured local sources have changed, while historical versions keep their original inputs. Older records without sources remain readable. • Improve the read-only Agent planning prompt to check existing implementations and justify new dependencies while preserving the full agreed scope and necessary tests. • Source history is stored locally. It records version inputs, not proof that every conclusion is supported or that acceptance checks passed. Capture the chat again to include new web messages. PRIVACY AND USER CONTROL • No background scraping. • Conversation capture begins only after an explicit user action. • Captured conversations and working drafts are stored locally in the current Chrome profile. • The selected organizer API receives the conversation only when you request organization. • Optional enhancements run only after you configure and enable the selected provider. Jev receives bounded conversation excerpts or structured requirements; local Laya processes enhancement input on your computer. • Providers never switch automatically. Local Laya requires the optional companion, your existing Python environment, and a complete local model directory. SpecsRelay does not download or install Laya or its models. • Choosing Laya makes only the enhancement step local; the main organizer still uses your separately configured API. • Data is not sent to a SpecsRelay server. • API keys are stored locally and are never injected into chatbot pages or sent to the coding agent. • No in-extension advertising, analytics, telemetry, or sale of user data. REQUIREMENTS An API key for a supported organizer provider or compatible endpoint is required. SpecsRelay can hand off by opening a supported desktop coding agent or copying the reviewed prompt. The optional SpecsRelay local companion can enhance installed-agent detection and project-directory handling on macOS or Windows, but it is not required for the core browser workflow. SpecsRelay is designed for developers and teams who clarify requirements with AI and want a controlled, reviewable handoff to local coding tools.