Planu Feature Parity Matrix
Comparison of Planu surfaces across Claude Code, Gemini CLI, and Codex. Last updated: 2026-04-21 (SPEC-591 / SPEC-592).
Legend
| Symbol | Meaning |
|---|---|
| ✅ | Full support — native |
| ⚠️ | Partial / workaround needed |
| ❌ | Not supported |
| 🔄 | Fallback: single-turn inline execution |
Core SDD Workflow
| Planu Surface | Claude Code | Gemini CLI | Codex | Fallback Strategy |
|---|---|---|---|---|
create_spec | ✅ | ✅ | ✅ | — |
challenge_spec | ✅ | ✅ | ✅ | — |
check_readiness | ✅ | ✅ | ✅ | — |
update_status | ✅ | ✅ | ✅ | — |
validate | ✅ | ✅ | ✅ | — |
list_specs | ✅ | ✅ | ✅ (scoped) | Per-user path scoping applied |
search_specs | ✅ | ✅ | ✅ (scoped) | Per-user path scoping applied |
heal_spec_docs | ✅ | ✅ | ✅ | — |
Estimation & Planning
| Planu Surface | Claude Code | Gemini CLI | Codex | Fallback Strategy |
|---|---|---|---|---|
estimate | ✅ | ✅ | ✅ | — |
plan_mode | ✅ | ✅ | 🔄 | No persistent multi-turn; inline |
decompose_spec | ✅ | ✅ | ✅ | — |
critical_path_analyzer | ✅ | ✅ | ✅ | — |
Orchestration & Agents
| Planu Surface | Claude Code | Gemini CLI | Codex | Fallback Strategy |
|---|---|---|---|---|
execute_sdd_flow | ✅ | ✅ | 🔄 | No Agent Teams; inline execution |
agent_swarm_status | ✅ | ⚠️ | 🔄 | No persistent subagent state |
multi_agent_review | ✅ | ⚠️ | 🔄 | Single-agent inline review |
orchestrate_runtime | ✅ | ⚠️ | 🔄 | Sequential inline execution |
Project Intelligence
| Planu Surface | Claude Code | Gemini CLI | Codex | Fallback Strategy |
|---|---|---|---|---|
init_project | ✅ | ✅ | ✅ | Codex: also scaffolds .openai/ |
detect_project_dna | ✅ | ✅ | ✅ | — |
bootstrap_project_intelligence | ✅ | ✅ | ✅ | — |
suggest_tooling | ✅ | ✅ | ✅ | — |
Audit & Compliance
| Planu Surface | Claude Code | Gemini CLI | Codex | Fallback Strategy |
|---|---|---|---|---|
export_audit_trail | ✅ | ✅ | ✅ | Codex: includes RBAC decisions |
check_compliance | ✅ | ✅ | ✅ | — |
generate_compliance_report | ✅ | ✅ | ✅ | — |
| RBAC gate | ❌ (n/a) | ❌ (n/a) | ✅ | SSO passthrough identity |
| Per-user path scoping | ❌ (n/a) | ❌ (n/a) | ✅ | Multi-user workspace isolation |
Configuration & Hooks
| Planu Surface | Claude Code | Gemini CLI | Codex | Fallback Strategy |
|---|---|---|---|---|
| Config scaffold | CLAUDE.md / .claude/ | GEMINI.md / .gemini/ | AGENTS.md / .openai/ | Host-specific |
| Hook system | Husky / post-checkout | Gemini hooks (TOML) | MCP config.toml | Host-specific |
configure_hooks | ✅ | ✅ | ✅ (config.toml) | — |
reconcile_skills | ✅ | ✅ | ✅ | — |
MCP Transport
| Planu Surface | Claude Code | Gemini CLI | Codex | Notes |
|---|---|---|---|---|
| stdio transport | ✅ | ✅ | ✅ | All three support native MCP-over-stdio |
| Streamable HTTP | ✅ | ⚠️ | ✅ | Codex natively supports both |
| REST/OpenAPI shim | n/a | n/a | ❌ not needed | Decision gate: MCP sufficient |
Notes
Codex path scoping:
list_specsandsearch_specsautomatically scope results to<workspaceRoot>/<tenantId>/<userId>/whenOPENAI_ENTERPRISE_TENANTis set. Single-user or non-enterprise Codex deployments have no scoping applied.Fallback tools: Tools marked 🔄 execute inline in a single turn. No output is lost; multi-agent features degrade gracefully to single-agent equivalents. Every fallback event is logged to
audit-trail.jsonl.Language stack: All Planu surfaces are language-agnostic — they operate on MCP protocol and file metadata, not source code. Parity is identical across TypeScript, Python, Go, Rust, Java, Swift, PHP, Ruby, C#, and Dart projects on all three hosts.