Codex Roadmap
This is the active implementation roadmap for the `codex/*` branch family.
This is the active implementation roadmap for the codex/* branch family.
Product Truth
MindStore should be usable in three practical ways:
- Self-hosted with PostgreSQL and no hosted AI provider
- Self-hosted or hosted with API-key providers
- Eventually, provider-account login for users who want to use their existing AI subscriptions instead of direct API billing
Near-Term Agenda
1. Provider Access Foundation
- Make requirements and auth modes explicit in product surfaces and docs
- Introduce a provider-access abstraction instead of assuming API keys forever
- Prepare encrypted auth-profile storage for future subscription-style provider login
2. MCP Modernization
- Migrate
/api/mcptoward the official MCP TypeScript SDK - Keep MindStore's current tools working during the migration
- Register plugin-provided tools, resources, and prompts through the runtime rather than hardcoded lists
3. Plugin Runtime Deepening
- Make runtime/plugin surfaces first-class in MCP
- Reduce drift between plugin metadata and actual execution surfaces
- Keep built-ins and community plugin surfaces aligned
4. Jobs and Scheduling
- Move plugin jobs beyond manual execution
- Add scheduling and execution groundwork for recurring or queued runtime jobs
- Keep the design compatible with a future worker service
5. MindStore Everywhere
- Improve site adapters and capture quality
- Keep packaging and setup simple for non-technical users
- Evolve hosted auth ergonomics after the provider-auth foundation exists
Integration Principle
Feature-heavy work from other branches should be integrated after the codex foundation is strong enough to absorb it cleanly.
Branch Convergence
The active convergence strategy is:
- preserve
fraindocs and planning artifacts that Irfan actively uses - use
codexas the architectural destination branch - port
frainfeature plugins into the codex runtime shape one by one - start with
kindle-importer,flashcard-maker, andvoice-to-memory