Why it matters
Agents start each session knowing nothing about how your week has gone. If they could see that you've been switching projects all day, or that one repository keeps producing failed runs, they could plan work around it. That context already exists in your tool history; agents can't read it without help.
How Ascenda helps
Flow runs a local MCP server with two tools, one for work sessions and one for working days. An agent can read:
- start and end times, the tool and model used, and prompts and minutes per session
- hands-on and supervising time, and how many sessions ran at once
- stopped runs, failed tool calls, compactions and tokens
- an hour-by-hour view of each day
Projects come back as opaque digests rather than names or paths, and the agent is told which one it's working in. The tools answer locally, only to agents you've connected.