Wispr Flow: Dictation That Works EverywhereStop typing. Start speaking. 4x faster.
Promoted
jobctl
The job application tracker your AI agent keeps updated
46 followers
The job application tracker your AI agent keeps updated
46 followers
jobctl is a job application tracker built MCP-first: AI agents (Claude, ChatGPT, Cursor) are the primary users - they record applications, interviews, salaries and next actions while you talk to them. You review everything in a clean dashboard: pipeline board, analytics, interview timeline, CV attachments. Remote MCP server with OAuth 2.1, listed in the official MCP Registry. Free.
Day 2 update from the maker - and a question I genuinely want your take on.
First: thank you for the warm welcome, the comments and the first users - watching real MCP connections appear in the logs never gets old.
The question: where should an AI agent's autonomy END in a tool like this? Adding an application from "I applied to Bosch, prio A, follow up Friday" is clearly fine. Deleting data is clearly not. But the middle is interesting: should an agent be allowed to mark an application as rejected on its own? Update your salary expectation? Draft a follow-up email? In jobctl the deterministic core validates every write and everything lands in an audit feed - but the permission line itself is a design choice, and I want to draw it where users actually want it. Tell me where you would draw it.