Problem
Professional agent stacks such as Hermes and OpenClaw let users define persona, long-term memory, and working agreements in portable Markdown files (e.g. Agent.md, Memory.md). Those files are versionable, backup-friendly, and easy to carry across machines or repos.
Grok Bot today does not offer an equivalent first-class mechanism. Persona, durable preferences, and standing instructions live mostly in opaque product state, which is harder for power users to review, diff, export, or share as a reproducible agent setup.
Proposal
Please add support for user-authored Markdown configuration that Grok Bot can load and apply, for example:
- Agent.md (or equivalent) — define the agent’s persona, responsibilities, constraints, and tool preferences in Markdown.
- Memory.md (or equivalent) — maintain long-term memory / preferences that persist across sessions and can be exported or backed up.
- File-based loading — ideally, placing these files in a repo or a local path is enough for the bot to read them and take effect (similar to how Hermes / OpenClaw work for professional users).
Why this matters
- Gives advanced users a clear, portable contract for how their bot should behave.
- Makes agent setup reviewable in git, shareable with teammates, and recoverable after resets.
- Aligns Grok Bot with workflows already common among serious agent users.
Scope notes
- Primary ask: Markdown-defined agent + memory that Grok Bot can load.
- Product: Grok Bot / Feature Requests.
- Happy to discuss exact filenames, discovery paths, and precedence vs in-app settings.
Thanks!