Pick the model path
Choose supported local, LAN, provider-key or Verdict Cloud routes. Model inference and tool execution can run in different places.
Verdict Code / terminal coding agent
Fix the bug.
See the evidence.
Choose a supported model. Work in your terminal. Review the change, the checks and the evidence that came back.
Local · LAN · BYOK · Verdict Cloud
Model and service readiness apply.
Example commands · ~/your-project
❯ verdict-code› /settingsFind the controls for the way you work.The Verdict difference
More control over the work.
More context for the review.
Check the outcome
Opt-in Verdict Proof Mode adds checks, repair and a structured proof result, with incomplete or unavailable evidence disclosed.
Read the current limits →Bring another view
Opinion presents separate responses and a synthesis. Fusion offers a gated fan-out/merge workflow. Neither guarantees correctness.
37 features / 4 agents
The comparison at a glance.
Verdict first. Every feature in one compact view. Select a feature to jump to its conditions and sources.
Supported capability Conditional / limited Unknown Row winner / tie
Colors show capability coverage, not measured quality. Unknown does not mean absent. Reviewed 2026-10-13; Verdict source baseline 1.4.537. Consumer verification is partial; source presence is not a release-availability claim.
| Feature | Verdict Code | Claude Code | Codex CLI | Pi |
|---|---|---|---|---|
| Interactive terminal | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| One-shot / piped input | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| File read/search/edit | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Shell / tests | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Local model inference | Verdict Code: Supported capability | Claude Code: Unknown from inspected evidence | Codex CLI: Supported capability | Pi: Supported capability |
| LAN model inference | Verdict Code: Conditional or limited scope | Claude Code: Unknown from inspected evidence | Codex CLI: Conditional or limited scope | Pi: Conditional or limited scope |
| BYOK / provider choice | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Offline coding | Verdict Code: Conditional or limited scope | Claude Code: Unknown from inspected evidence | Codex CLI: Conditional or limited scope | Pi: Conditional or limited scope |
| Model selection / pinning | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Permission controls | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Conditional or limited scope |
| OS sandbox | Verdict Code: Conditional or limited scope | Claude Code: Conditional or limited scope | Codex CLI: Supported capability; coverage winner or tie | Pi: Conditional or limited scope |
| Plan before edits | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Conditional or limited scope |
| Web search / fetch | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Conditional or limited scope |
| Image inputs | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| MCP client | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Skills / custom commands | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Lifecycle hooks | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Background agents | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Conditional or limited scope |
| Resume session | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Branch conversation | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Local transcript export | Verdict Code: Supported capability | Claude Code: Supported capability | Codex CLI: Unknown from inspected evidence | Pi: Supported capability |
| Undo / recovery | Verdict Code: Conditional or limited scope | Claude Code: Conditional or limited scope | Codex CLI: Conditional or limited scope | Pi: Conditional or limited scope |
| Context compaction | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Persistent instructions / memory | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Usage / cost visibility | Verdict Code: Conditional or limited scope | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Budget controls | Verdict Code: Conditional or limited scope | Claude Code: Conditional or limited scope | Codex CLI: Unknown from inspected evidence | Pi: Unknown from inspected evidence |
| Verdict Proof Mode | Verdict Code: Conditional or limited scope | Claude Code: Conditional or limited scope | Codex CLI: Unknown from inspected evidence | Pi: Unknown from inspected evidence |
| Second opinion / consultation | Verdict Code: Conditional or limited scope | Claude Code: Conditional or limited scope | Codex CLI: Conditional or limited scope | Pi: Conditional or limited scope |
| Multi-model Fusion | Verdict Code: Conditional or limited scope | Claude Code: Unknown from inspected evidence | Codex CLI: Unknown from inspected evidence | Pi: Conditional or limited scope |
| Verdict Decision | Verdict Code: Conditional or limited scope | Claude Code: Unknown from inspected evidence | Codex CLI: Unknown from inspected evidence | Pi: Conditional or limited scope |
| Machine-readable output | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| Git / PR workflows | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
| IDE integration | Verdict Code: Conditional or limited scope | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Conditional or limited scope |
| Cloud execution | Verdict Code: Unknown from inspected evidence | Claude Code: Supported capability | Codex CLI: Supported capability | Pi: Unknown from inspected evidence |
| Mobile / remote continuation | Verdict Code: Conditional or limited scope | Claude Code: Conditional or limited scope | Codex CLI: Conditional or limited scope | Pi: Conditional or limited scope |
| Recurring work | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Conditional or limited scope | Pi: Conditional or limited scope |
| Slash discoverability / settings | Verdict Code: Supported capability; coverage winner or tie | Claude Code: Supported capability; coverage winner or tie | Codex CLI: Supported capability; coverage winner or tie | Pi: Supported capability; coverage winner or tie |
What a highlighted winner means
A row winner has the strongest established coverage of that row’s named capability, with fewer listed scope limits. Equal coverage is a tie. If a competing capability is unknown, no winner is declared. Read the row criterion and the conditions before choosing.
Verdict entries come from source inspection. Claude Code, Codex CLI and Pi entries come from official documentation. These are editorial capability judgments; the agents were not tested head-to-head for accuracy, speed or cost. Account, model, platform and feature gates affect availability.
The full picture
Every feature. Conditions included.
Read the exact scope behind each color. Competitor sources link directly to official documentation.
| Feature / coverage criterion | Verdict Code | Claude Code | Codex CLI | Pi |
|---|---|---|---|---|
| Interactive terminalBuilt-in or documented support for interactive terminal. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner Implemented terminal loop. | Supported capabilityTied winner claude interactive session. | Supported capabilityTied winner codex interactive TUI. | Supported capabilityTied winner Interactive terminal UI. |
| One-shot / piped inputBuilt-in or documented support for one-shot / piped input. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner -1, -p, --prompt-file, --prompt-stdin. | Supported capabilityTied winner claude -p; stdin pipelines. | Supported capabilityTied winner codex exec; prompt plus piped context. | Supported capabilityTied winner Print mode and piped input. |
| File read/search/editBuilt-in or documented support for file read/search/edit. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner Read, Glob, Grep, Write, Edit; deferred tool discovery. | Supported capabilityTied winner Reads code, edits files, runs commands. | Supported capabilityTied winner Inspects files, edits code, runs installed development tools. | Supported capabilityTied winner Built-in read, edit, write, grep and find tools. |
| Shell / testsBuilt-in or documented support for shell / tests. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner Bash, background execution, task tools; model/tool capability dependent. | Supported capabilityTied winner Shell tools and test execution. | Supported capabilityTied winner Shell execution and local development tools. | Supported capabilityTied winner Bash tool; background shell work uses tmux. |
| Local model inferenceBuilt-in or documented support for local model inference. Conditions apply as listed.No declared winner | Supported capability Local provider implementation; model/backend dependent. | Unknown from inspected evidence Native arbitrary local-model/offline support unknown from inspected docs; compatible gateways documented. | Supported capability --oss supports Ollama and LM Studio. | Supported capability Local llama.cpp router and other supported local endpoints. |
| LAN model inferenceBuilt-in or documented support for lan model inference. Conditions apply as listed.No declared winner | Conditional or limited scope Explicit lan/ provider, endpoint discovery and hostname handling. Campaign LAN route remains NOT_RUN. | Unknown from inspected evidence Gateway base-URL overrides documented; arbitrary LAN inference support unknown. | Conditional or limited scope Custom provider base_url allows compatible network endpoints; LAN discovery/catalog behavior unknown. | Conditional or limited scope Custom provider endpoints; automatic LAN discovery is not established. |
| BYOK / provider choiceBuilt-in or documented support for byok / provider choice. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner Local, LAN, Direct BYOK, Gateway and Verdict Credits routing implementations. | Supported capabilityTied winner Anthropic plus supported AWS, Google and Microsoft deployments. | Supported capabilityTied winner API credentials, custom providers, proxies, Azure configuration and built-in Bedrock provider. | Supported capabilityTied winner Multiple providers, API keys, OAuth and custom providers. |
| Offline codingBuilt-in or documented support for offline coding. Conditions apply as listed.No declared winner | Conditional or limited scope Possible with ready local models and local tools; service-dependent features cannot be assumed offline. | Unknown from inspected evidence Fully offline coding unknown from inspected official docs. | Conditional or limited scope Local OSS inference documented; fully disconnected end-to-end workflow not tested here. | Conditional or limited scope Local inference plus offline network controls; disconnected coding not tested here. |
| Model selection / pinningBuilt-in or documented support for model selection / pinning. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner -m; Auto selection with source scope/performance profile; explicit-pin failover handling. | Supported capabilityTied winner --model, model aliases and organization allowlists. | Supported capabilityTied winner Model selection, custom providers and saved profiles. | Supported capabilityTied winner Model flags, /model and scoped model cycling. |
| Permission controlsBuilt-in or documented support for permission controls. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI | Supported capabilityTied winner Safe/ask/full modes, per-tool approval, path/domain boundaries. | Supported capabilityTied winner Permission modes and tool rules. | Supported capabilityTied winner Approval policies, rules, writable roots and /permissions. | Conditional or limited scope Project trust; per-tool approval requires an extension or external controls. |
| OS sandboxOS-enforced isolation for built-in shell execution across supported platforms. This does not compare MCP, hooks or other external tools; narrow shell boundaries or external setup count as conditional.Codex CLI | Conditional or limited scope Seatbelt exists for portable workspace Bash and specific proof oracles. General permission gates do not establish universal process isolation. | Conditional or limited scope Shell sandbox on macOS/Linux/WSL2; off by default; file tools, MCP and hooks outside it. Native Windows commands unsandboxed. | Supported capabilityRow winner OS-enforced shell sandbox and writable-root controls; read-only/workspace-write/full-access modes. Do not assume MCP servers or other external processes inherit the same boundary. | Conditional or limited scope External container or sandbox; Pi tools otherwise inherit process permissions. |
| Plan before editsA built-in mode that prevents edits while the agent plans.Tie: Verdict Code, Claude Code, Codex CLI | Supported capabilityTied winner /plan on, off or status controls the existing mutating-tool gate. A user lock prevents the agent from leaving plan mode. This is a tool permission boundary, not an OS sandbox. | Supported capabilityTied winner Plan mode reads/proposes without editing. | Supported capabilityTied winner /plan switches to plan mode. | Conditional or limited scope Plan workflow available through extensions or packages; no built-in plan mode. |
| Web search / fetchBuilt-in or documented support for web search / fetch. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI | Supported capabilityTied winner WebSearch/WebFetch implementations. /search searches repository documentation, not the internet. | Supported capabilityTied winner WebSearch/WebFetch tools documented. | Supported capabilityTied winner Cached search and live --search; custom-provider support can differ. | Conditional or limited scope Can add search/fetch through MCP or extensions; dedicated built-in tools not established. |
| Image inputsBuilt-in or documented support for image inputs. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner Repeatable --image and /image; model vision support required. | Supported capabilityTied winner Image paste, drag/drop or file-path reference. | Supported capabilityTied winner Image paste or --image; PNG/JPEG documented. | Supported capabilityTied winner Image file inputs; requires a model with image support. |
| MCP clientBuilt-in or documented support for mcp client. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner MCP config, registry/client and /mcp; service/server dependent. | Supported capabilityTied winner Local stdio and remote HTTP/SSE/WebSocket configurations. | Supported capabilityTied winner Stdio/streamable HTTP, OAuth, required-server and tool controls. | Supported capabilityTied winner Built-in MCP, stdio/HTTP, OAuth and deferred tool discovery. |
| Skills / custom commandsBuilt-in or documented support for skills / custom commands. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner Skill loading, manager/authoring commands and slash discovery. | Supported capabilityTied winner SKILL.md skills; custom commands integrated into skills. | Supported capabilityTied winner Reusable skills plus installable plugins/MCP bundles. | Supported capabilityTied winner Skills, prompt templates, extensions and installable packages. |
| Lifecycle hooksBuilt-in or documented support for lifecycle hooks. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner SessionStart/End, Pre/PostToolUse, PostToolUseFailure, UserPromptSubmit and PreCompact implemented. | Supported capabilityTied winner Lifecycle hooks including PreToolUse documented. | Supported capabilityTied winner Command/MCP hooks, including tool, session, compaction and subagent events; non-managed hooks require trust review. | Supported capabilityTied winner Extension events for sessions, model calls, tools and compaction. |
| Background agentsBuilt-in managed subagent/background agent workflows, rather than user-authored extensions.Tie: Verdict Code, Claude Code, Codex CLI | Supported capabilityTied winner Task lifecycle, background tracker and /agents --view. | Supported capabilityTied winner Custom/background subagents and parallel work. | Supported capabilityTied winner Subagents, custom roles and /agent thread navigation. | Conditional or limited scope Subagent workflows through extensions or tmux; not built into the core. |
| Resume sessionBuilt-in or documented support for resume session. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner -c, -r, /resume, local persistence/journal. | Supported capabilityTied winner Continue/resume by ID or name. | Supported capabilityTied winner codex resume, exec resume. | Supported capabilityTied winner Continue/resume flags and /resume session picker. |
| Branch conversationA documented command that branches the conversation without treating Git branches as conversation forks.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner /fork [title] saves an independent conversation snapshot with parent identity. The current session continues; resume the copy in a new verdict-code --resume invocation. It does not create a Git branch. | Supported capabilityTied winner --fork-session, /branch. | Supported capabilityTied winner codex fork, /fork, edit-and-fork shortcut. | Supported capabilityTied winner /tree, /fork and /clone preserve conversation alternatives. |
| Local transcript exportExport the conversation to a local file through an explicit terminal command. Single-reply copying and inference event streams do not establish full-conversation export.No declared winner | Supported capability /export <path.json|path.md> writes the conversation locally inside the current workspace. Existing files are protected unless --overwrite is explicit. Source-inspected; broader consumer checks remain partial. | Supported capability /export [filename] exports the current conversation as plain text; a supplied filename writes directly to a file. | Unknown from inspected evidence A built-in full-conversation local export command is not established by the inspected docs. /copy copies a reply; JSON run events are a different scope. | Supported capability /export [path] writes the current session to HTML or JSONL. Separate sharing commands can upload a session; local export is distinct. |
| Undo / recoveryBuilt-in or documented support for undo / recovery. Conditions apply as listed.No declared winner | Conditional or limited scope /rewind alias /undo; tracked-change and Git fallback implementation. Not a universal rollback guarantee. | Conditional or limited scope Checkpoint rewind for tracked direct edits; Bash/external changes and most subagent edits excluded. | Conditional or limited scope Conversation edit/fork documented; equivalent automatic file checkpoint rewind unknown. Docs recommend Git checkpoints. , | Conditional or limited scope Conversation tree recovery; automatic file undo not established. |
| Context compactionBuilt-in or documented support for context compaction. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner /compact, auto-compaction, microcompaction, ledger/memory implementations. | Supported capabilityTied winner /compact and automatic compaction controls. , | Supported capabilityTied winner /compact summarizes session context. | Supported capabilityTied winner Automatic compaction and /compact with custom instructions. |
| Persistent instructions / memoryBuilt-in or documented support for persistent instructions / memory. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner VERDICT.md and session memory implementations; project AGENTS.md is used only when neither ./VERDICT.md nor ./.verdict/VERDICT.md exists. Existing Verdict-file precedence is preserved. | Supported capabilityTied winner CLAUDE.md and automatic memory. | Supported capabilityTied winner AGENTS.md and local memory controls, including /memories. , | Supported capabilityTied winner Project instructions and context files; long-term memory can use extensions. |
| Usage / cost visibilityBuilt-in or documented support for usage / cost visibility. Conditions apply as listed.Tie: Claude Code, Codex CLI, Pi | Conditional or limited scope /cost, /usage, usage files, provider/Gateway receipts; campaign exposed incomplete coverage. | Supported capabilityTied winner /usage shows tokens/cost estimates; authoritative billing provider-side. | Supported capabilityTied winner /usage account activity; structured runs emit usage counters; subscription/API pricing differs. , , | Supported capabilityTied winner /session shows token usage and cost; custom providers must report usage. |
| Budget controlsBuilt-in or documented support for budget controls. Conditions apply as listed.No declared winner | Conditional or limited scope Token ceilings, verification time/credit controls and bounded retries implemented; scope-specific. | Conditional or limited scope --max-budget-usd, --max-turns; estimated-spend cap can overshoot. | Unknown from inspected evidence Equivalent universal per-run dollar cap unknown from inspected docs. | Unknown from inspected evidence A universal per-run dollar budget cap is not established by the inspected docs. |
| Verdict Proof ModeBuilt-in or documented support for verdict proof mode. Conditions apply as listed.No declared winner | Conditional or limited scope Opt-in check/repair workflow, structured proof result and evidence/usage disclosure. C-S6 remains PARTIAL, not PASS. | Conditional or limited scope No verified equivalent to Verdict's specific receipt/terminal contract; /verify provides model-driven app verification. | Unknown from inspected evidence No verified equivalent to Verdict's specific receipt/terminal contract; coding/tests/review documented. | Unknown from inspected evidence An equivalent proof result and terminal evidence contract is not established. |
| Second opinion / consultationBuilt-in or documented support for second opinion / consultation. Conditions apply as listed.No declared winner | Conditional or limited scope /opinion: multi-model advisory synthesis, presets/models/Decision adjudication. Verdict Server dependent. | Conditional or limited scope Experimental /advisor consults a second model; provider/model eligibility restrictions apply. | Conditional or limited scope Multi-agent/model roles allow separate reviews; dedicated fixed Opinion-style product unknown. | Conditional or limited scope Custom extensions can orchestrate additional models; dedicated consultation UI not established. |
| Multi-model FusionBuilt-in or documented support for multi-model fusion. Conditions apply as listed.No declared winner | Conditional or limited scope /fusion: tool-free one-shot fan-out/merge; feature-gated; checks selected architecture; cloud-direct route unsupported. | Unknown from inspected evidence Arbitrary multi-provider fan-out/merge equivalent unknown; advisor/subagents offer related workflows. , | Unknown from inspected evidence Dedicated Fusion equivalent unknown; subagents offer related orchestration. | Conditional or limited scope Custom extensions can compose model calls; dedicated fan-out/merge product not established. |
| Verdict DecisionBuilt-in or documented support for verdict decision. Conditions apply as listed.No declared winner | Conditional or limited scope Small scoring-model features for logs, completion claims, stalls, risk, incidents and sizing; admission/readiness gates. Advice does not certify work. | Unknown from inspected evidence Dedicated equivalent portfolio unknown; advisor, hooks and review offer related mechanisms. , | Unknown from inspected evidence Dedicated equivalent portfolio unknown; subagents/hooks/review offer related mechanisms. , | Conditional or limited scope Classifier primitives are documented; an equivalent built-in Decision portfolio is not established. |
| Machine-readable outputBuilt-in or documented support for machine-readable output. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner Text/JSON/stream-json; --json-schema validation. | Supported capabilityTied winner JSON/stream-json and JSON Schema outputs. | Supported capabilityTied winner JSONL events and --output-schema. | Supported capabilityTied winner JSONL events and RPC; equivalent output-schema validation not established. |
| Git / PR workflowsGit and pull-request work through available shell/MCP tools with required Git utilities and authentication. No dedicated PR command is required.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner Built-in Git actions; shell/MCP can create PRs with the required utilities and authentication. | Supported capabilityTied winner Commits and PR workflows documented; available Git utilities and authentication required. | Supported capabilityTied winner Diff/review and shell tools; Git/PR work requires available utilities and authentication. | Supported capabilityTied winner Built-in shell tool can invoke Git/PR utilities; the required tools and authentication must be available. |
| IDE integrationBuilt-in or documented support for ide integration. Conditions apply as listed.Tie: Claude Code, Codex CLI | Conditional or limited scope /ide VS Code bridge for context, diffs and session sync. Implemented, not exercised here. | Supported capabilityTied winner VS Code/Cursor and JetBrains companion integrations. | Supported capabilityTied winner IDE companion extension; CLI /ide imports editor context. , | Conditional or limited scope SDK/RPC embedding documented; official companion IDE integration not established. |
| Cloud executionA documented CLI command that submits and tracks coding tasks in a managed cloud environment.No declared winner | Unknown from inspected evidence Cloud/provider routing exists. Full CLI-to-cloud coding-task dispatch lifecycle not established by this inspection. | Supported capability CLI cloud dispatch plus hosted/self-hosted cloud sessions. | Supported capability codex cloud submit/list and codex apply. | Unknown from inspected evidence Dedicated CLI-to-cloud coding-task dispatch lifecycle not established. |
| Mobile / remote continuationBuilt-in or documented support for mobile / remote continuation. Conditions apply as listed.No declared winner | Conditional or limited scope /remote-control requires Verdict Server, bridge daemon and cloud redemption. Installed cross-device flow not tested here. | Conditional or limited scope CLI Remote Control continues local sessions through browser/mobile companions. | Conditional or limited scope CLI remote-control commands documented; mobile companion setup starts in desktop app, with rollout/account gates. , | Conditional or limited scope Terminal/tmux and RPC integrations; equivalent mobile session handoff not established. |
| Recurring workA built-in CLI command for recurring prompts in the current session.Tie: Verdict Code, Claude Code | Supported capabilityTied winner /loop recurring prompts/slash commands; local session mechanics. | Supported capabilityTied winner /loop within CLI; /schedule cloud routines and desktop schedules have separate scopes. | Conditional or limited scope Script/CI automation documented; desktop scheduled tasks separate surface. , | Conditional or limited scope Scripts or extensions can schedule work; a built-in /loop is not established. |
| Slash discoverability / settingsBuilt-in or documented support for slash discoverability / settings. Conditions apply as listed.Tie: Verdict Code, Claude Code, Codex CLI, Pi | Supported capabilityTied winner /settings forwards registered controls; /settings search filters names, aliases, descriptions and categories without executing matches. Launch and terminal references describe new-invocation options; they do not change the active session. | Supported capabilityTied winner Menu lists commands/skills/plugins; some availability varies and some commands intentionally hidden. | Supported capabilityTied winner Slash popup; model, permissions, plan, hooks, MCP, skills/plugins, usage and session controls. | Supported capabilityTied winner Slash menu searches loaded commands; /settings and /model documented. |
No verdict without evidence.
What this comparison does—and does not—prove
Source presence is not live acceptance. The reviewed campaign left Verdict Proof Mode partial and LAN execution untested. BYOK/direct routing coverage, Gateway fallback, standalone overrides and local-pin test-author routing remain limitations in this baseline. This page does not claim complete audit coverage, fully local Proof Mode or every feature offline.
Verdict implementation sources and review boundary
Inspected Verdict Code 1.4.537, 2026-10-13. Source paths below are beneath tools/verdict-code/verdict_code/. They are implementation evidence, not performance results.
- CLI / settings: main.py; commands/builtin/settings.py; commands/builtin/plan.py; commands/builtin/transcript.py; commands/discovery.py; commands/router.py; agent.py.
- Tools / permissions: tools/catalog.py; tools/__init__.py; tools/permission_gate.py.
- Model routes: provider/registry.py; provider/local_provider.py; provider/lan_provider.py; provider/direct_byok_provider.py; model_policy.py.
- Extensions / agents: mcp/; skills/; hooks/; subagents/; tools/task.py.
- Context / recovery: session/manager.py; session/journal.py; memory/; commands/builtin/resume.py; commands/builtin/rewind.py.
- Evidence / multi-model workflows: smart_harness_runner.py; smart_harness_terminal_contract.py; commands/builtin/opinion.py; commands/builtin/fusion_cmd.py; decision_client.py.
- Companions / recurring work: commands/builtin/ide.py; ide/bridge.py; commands/builtin/remote_control.py; commands/builtin/loop.py.
Pi reflects the current Pi 1.0 documentation, including built-in MCP. An extension or external workflow is distinguished from a built-in feature. No unverified pricing, testimonials or performance rankings are used.