1. Keep Lex boring and dependable.
Lex remains the public memory and policy substrate: stable CLI behavior, useful MCP coverage, deterministic JSON, local SQLite, and better instructions generation.
Publication status • September 6, 2026
The roadmap is moving from tools to an operating stack.
Lex, AXF, LexSona and LexRunner have public source repositories and npm releases. AXF, LexSona and LexRunner are open source under Apache-2.0; Lex remains MIT. ContextForge is a proposed context linker. LexThority is historical research with no active product commitment. Behavior and containment boundaries continue to evolve.
The near-term work is not another prompt pattern. It is the layer that lets agents remember, discover workspace capabilities, ask for authority, inherit behavior, and coordinate bounded work.
Current public surface
| Project | Status | Package | Source |
|---|---|---|---|
| Lex | Public MIT core for memory, policy, recall, instructions, CLI, MCP, and API. | npm latest | GitHub |
| Lex MCP | Public MCP wrapper for Lex surfaces. | npm latest | NPM package |
| AXF | Public Apache-2.0 framework for workspace capability contracts. | npm latest | GitHub |
| ContextForge | Proposed pure context linker; ownership and input contracts await ratification. | No npm release or executable implementation. | Proposal page / Public decision |
| LexThority | Historical authority research; no active product commitment. | No npm release. | GitHub |
| LexSona | Public Apache-2.0 behavioral constraint engine. | npm 2.0.4 | GitHub |
| LexRunner | Public Apache-2.0 flow layer for gates and receipts, with containment and coordination under development. | npm 2.4.0 | GitHub |
Near-term roadmap
2. Make AXF the workspace capability layer.
AXF should make local repo know-how discoverable and runnable without turning every capability into raw shell improvisation. Expect more adapter examples and lifecycle hardening.
3. Ratify ContextForge and its input contracts.
Confirm ownership and exact released Lex and domain contracts before implementing the pure context linker. Deterministic bundles and explanations come first; a LexRunner consumer follows. Authentication and effect authorization remain independent work.
4. Separate LexSona behavior from LexRunner flow.
LexSona should return deterministic behavioral constraints. LexRunner should consume those constraints, not become the persona engine.
5. Add contained swarm execution to LexRunner.
Parallel agents need contained, worktree-like execution roots, per-agent envelopes, explicit mounts, and gated promotion back into canonical work.
6. Prove the stack outside development.
Analyst, product, support, security, and operations workflows should use the same surfaces: memory, capability, authority, behavior, flow, and receipts.
Design principles
Use existing brokers
The stack should cooperate with OAuth, cloud IAM, GitHub permissions, and enterprise security tooling. It should not invent a shadow identity provider.
Fail with operator-ready detail
When a credential, policy, or tool blocks work, the agent should be able to tell an operator what was attempted, what failed, and what permission or action is needed.
Preserve agent autonomy
The goal is bounded agency, not bureaucracy. Agents should reason freely inside a workspace while effects are gated, receipted, and recoverable.