Roadmap
What's coming next in grith.
The roadmap is necessarily speculative — dates slip, priorities shift. But here's the public-facing plan as of v0.1.0 launch.
v0.1.x — stabilisation (Q3 2026)
- Profile drift detection continuous monitoring.
- Profile signing UX improvements.
- Performance — squeeze the secret scanner; aim for p99 under 12ms.
- Audit log integrity — append-only / WORM option.
v0.2 — platform expansion (Q4 2026)
- aarch64 Linux — register backend + seccomp-bpf for ARM.
- Custom filter SDK — open-source Rust API for authoring user filters, shipped via Pro policy bundle.
- Versioned docs UI — version selector on docs.grith.ai.
- Reproducible builds — bit-for-bit reproducible release artefacts.
v1.5 — semantic analysis (Q1 2027)
- Filter 17 fully implemented — local embedding model, intent classification.
- Cross-session taint analytics — Pro feature.
- Behavioural baseline export — share baselines across team members optionally.
v2.0 — multi-platform (Q2-Q3 2027)
- macOS supervisor — Endpoint Security framework.
- Windows supervisor — ETW + minifilter.
- SSO / SAML — Okta, Azure AD, Google Workspace.
- RBAC — full role-based access control on the dashboard.
- Air-gapped Enterprise — offline license activation, self-hosted dashboard, manual policy distribution.
- Discord bot integration — interactive approve/deny in Discord (parity with Slack/Telegram).
- Microsoft Teams interactive cards — parity for Adaptive Cards.
v2.x — broader scope
- Browser-extension supervisor — for browser-based agents.
- Mobile agent supervision — for the inevitable mobile agent runtimes.
- Federated reputation — opt-in cross-organisation trust sharing for enterprise consortiums.
How to influence
The roadmap is informed by community feedback. The strongest signal:
- GitHub issues with the
enhancementlabel, sorted by reactions. - GitHub Discussions for design conversations.
- Pro / Enterprise customer requests — taken seriously, but balanced against the open-source community's needs.
We don't bias toward paying customers' wishes on architectural decisions; we do bias toward them on prioritisation of work already-planned.
Things we will not do
- Tie the community edition to a grith.ai service. The community edition is and remains usable offline indefinitely.
- Add LLM-based decision-making to the filter pipeline. Filters stay deterministic.
- Add a marketplace. Filters are policy; we won't run a "Filter Store".
See also
Last updated: 2026-05-14Edit this page on GitHub →