Why now?

Agent-switching friction is a live discussion topic right now: the HN thread drew 68 points and 62 comments, and an autocomplete signal of 16 suggests organic search interest in the problem. Notion is shipping this as a feature of its existing workspace, which signals the problem is real but also that the window for a standalone product is narrow.

The gap

The gap is cross-tool context portability. Notion's version only works because it already owns the user's workspace data; knowledge workers who spread work across Notion, Google Docs, GitHub and local files still re-explain themselves when switching agents. No signal here proves anyone will pay for a neutral layer — demand is discussed, not proven.

Main competitor

Notion itself, plus agent vendors (OpenAI, Anthropic, Cursor) absorbing context persistence as a free built-in feature.

Execution plan

1. Build a thin local context store (a file or SQLite DB) that captures project state, decisions and open questions, and exports a compact briefing prompt for any agent — start with Claude, ChatGPT and Cursor.
2. Ship a CLI plus a single 'brief me' command that assembles the briefing from the store, so switching agents costs one paste instead of a re-explanation.
3. Validate with the HN/Reddit crowd discussing agent context loss: post the tool, measure whether people actually use it across multiple agents or just once.
4. Add connectors one at a time (local files, GitHub, then Notion export) so the tool stays neutral rather than tied to one platform.
5. If repeat usage shows up, gate multi-project history and team-shared briefings behind a paid tier; if not, keep it free and stop.

Monetization

Freemium: free single-project local context store, paid tier for multi-project history, team-shared briefings and hosted sync. No pricing or revenue figures are supported by the signals — treat willingness to pay as unproven until the validation step shows repeat cross-agent usage.