The README is explicit: do not add `"type": "stdio"` to the client entry. Variable substitution happens before type validation, so a client that sets it can end up resolving `${workspaceFolder}` into a nonsense path and the server never starts. Beyond that, the design decision worth making up front is storage mode — per-project keeps each repo's history isolated with no configuration, centralised is what you want if you care about comparing across projects.
A time and activity server for AI coding agents. It answers what time it is now, converts between timezones, and keeps a persistent log of activities — started, ended, durations, outcomes — so an agent can resume work in a later session and compare estimates against what actually happened.
- `get_current_time` takes a `timezone`; pass `"system"` or `"local"` for the machine's own time, or an IANA name like `"Europe/London"`
- `convert_time` translates a `time` between a `source_timezone` and a `target_timezone`, handling DST
- `start_activity_log` opens an activity with a type, scope, description and optional tags, and returns an Activity ID
- `end_activity_log` closes it with a result and notes, calculating duration automatically
- `get_elapsed_time` checks progress on a running activity without interrupting it
- `get_activity_logs` queries history with filtering, and `update_activity_log` corrects or annotates a finished entry
- `create_time_reminder` and `check_time_reminders` schedule and surface follow-ups
- Three Activity ID formats: `custom` at 12 characters, `short` at 22, `uuid` at 36
Python 3.10 or higher and an MCP-capable client. Published as chronos-protocol version 1.6.0; the documented path is clone, `pip install -r requirements.txt`, then `pip install -e .` — the editable install is required for MCP. No account or key. Storage is either per-project, writing to `{project-root}/chronos-data/time_server_data.json` with `--storage-mode per-project`, or centralised in a directory given by `--data-dir`. Launch as `python -m chronos_protocol` with those flags.
Build from source — clone the repository and build it, then point your client at the binary
