Finish Line
Every Code has a reviewable Codex-base substrate migration plan that replaces the raw spike with scoped PR slices, clarifies the future role of code-rs and codex-rs, and prevents new work from landing on the wrong base.
Current Status
State: Rescoped from "Codex-base substrate migration for Every Code" to "Codex CLI fork foundation plus narrow Every Code overlays." The substrate work landed; the active risk is stale planning/docs language that still sounds product-first.
Decision update, 2026-06-06:
- Codex CLI compatibility is the default direction for app-server/session/protocol/config/TUI behavior.
- Every Code is moving into maintenance mode as a standalone product direction.
- The repo should be treated as a Codex CLI fork that layers selected Every Code capabilities on top only when they remain needed and are validated.
code-rs is the editable fork implementation; codex-rs remains the read-only upstream provenance mirror until a new issue changes that role.
- Fork deltas should be small, reviewable, and easy to rebase/refresh from upstream Codex CLI.
Landed foundation:
Remaining alignment work:
Next action:
- Track the docs/policy cleanup as a small explicit issue so future agents stop inheriting the wrong north star.
Acceptance Criteria
- Decide and document the
code-rs / codex-rs roles for the Codex-base migration.
code-rs can build from a Codex-base workspace as the code binary with ./build-fast.sh.
- Imported upstream crates keep
codex-* names unless there is a specific Every Code product reason to alias/rename.
- Every Code-only additions have a documented
code-* port strategy.
CODE_HOME takes precedence over CODEX_HOME; no-env default is ~/.code.
- No
code-rs crate depends on sibling ../codex-rs; the path-dependency guard enforces the chosen policy.
.github/github.json, AGENTS/repo guidance, and docs/upstream-import-policy.md describe the new substrate/mirror roles.
- Active issues that touch app-server/session/protocol/TUI/remote-inbox/release identity are linked, blocked, or marked independent.
- The raw spike is decomposed into reviewable PR slices and is not merged wholesale.
Relationships
Finish Line
Every Code has a reviewable Codex-base substrate migration plan that replaces the raw spike with scoped PR slices, clarifies the future role of
code-rsandcodex-rs, and prevents new work from landing on the wrong base.Current Status
State: Rescoped from "Codex-base substrate migration for Every Code" to "Codex CLI fork foundation plus narrow Every Code overlays." The substrate work landed; the active risk is stale planning/docs language that still sounds product-first.
Decision update, 2026-06-06:
code-rsis the editable fork implementation;codex-rsremains the read-only upstream provenance mirror until a new issue changes that role.Landed foundation:
code-rswith the Codex-base Rust workspace while keepingcodex-rsas the read-only mirror and preservingCODE_HOME > CODEX_HOME > ~/.code.Remaining alignment work:
just-every/codeas fork/import history where relevant, but do not let it drive new product direction unless a current overlay issue says so.Next action:
Acceptance Criteria
code-rs/codex-rsroles for the Codex-base migration.code-rscan build from a Codex-base workspace as thecodebinary with./build-fast.sh.codex-*names unless there is a specific Every Code product reason to alias/rename.code-*port strategy.CODE_HOMEtakes precedence overCODEX_HOME; no-env default is~/.code.code-rscrate depends on sibling../codex-rs; the path-dependency guard enforces the chosen policy..github/github.json, AGENTS/repo guidance, anddocs/upstream-import-policy.mddescribe the new substrate/mirror roles.Relationships
spike/codex-base-portproved a Codex-base workspace could be assembled. It is no longer the active source of truth after PR docs: record Codex substrate migration policy #389/feat: replace Rust substrate with Codex base #390 and may be deleted once this note is sufficient evidence.