HarnessMatch
Menu

Coding agent

Oh My Pi

A batteries-included Pi fork with LSP, debugger, browser, and subagents.

Status
Active
License
MIT
Evidence
Documented + code-verifiable, 18 sources
Product record checked
2026-07-28
Compare harnessesInspect sources

At a glance

An MIT-licensed terminal agent, invoked as `omp`, that layers LSP and debugger tooling, browser control, first-class subagents, MCP, ACP, RPC, and a large provider catalog onto Pi's extensible base.

Good choice if

  • Terminal users who want IDE-grade code intelligence and debugger access
  • Parallel tasks with typed subagent results and isolated worktrees
  • Provider-flexible automation through one-shot, RPC, ACP, and SDK entry points

Check before choosing

  • The default approval mode is yolo, destructive-command overrides do not force a prompt in that mode, and subagents also run yolo; safer always-ask/write and per-tool policies require explicit configuration
  • Task copy-on-write or worktree isolation separates delegated file changes but is not a documented OS-level sandbox for the primary process, browser, debugger, shell, or local MCP servers
  • Checkpoint and rewind are disabled by default and prune conversation context only: they do not restore working-tree files, Git state, artifacts, processes, or external side effects
See 4 more considerations
  • Cross-session memory is disabled by default and model-generated; it must be checked against the current repository rather than treated as authoritative state
  • Secret obfuscation is disabled by default, while extensions and hooks run in-process without isolation and can block or rewrite tool calls, access session state, or terminate the session if they fail outside guarded handlers; revised inputs are schema-revalidated and pass through the approval gate before execution
  • The large built-in surface (browser/CDP, debugger, eval kernels, SSH, MCP, collaboration, media, and high-concurrency agents) expands setup and trust boundaries
  • Project-owned compaction and local-model evaluations plus README performance claims are not independent harness benchmarks, so no score is imported

Why it qualifies as a coding harness

This confirms category fit, not product quality. Every required criterion links back to first-party evidence.

Qualifies4 of 4 required criteria evidenced

  • Adaptive agent loop

    Documented

    The system repeatedly observes results and chooses the next action instead of following a fixed one-pass graph.

    EvidenceProduct overview

  • Repository tool execution

    Documented

    The system can use tools to inspect and change a repository or its execution environment.

    EvidenceTask subagents

  • Task-aware context management

    Documented

    The runtime assembles, updates, compacts, retrieves, or persists task-relevant context while work proceeds.

    EvidenceSettings reference

  • Model-independent runtime control

    Documented

    Permissions, budgets, interruption, policy, or stop controls operate outside the model's own text generation.

    EvidenceApproval modes

Membership establishes category fit only. It does not score quality, safety, autonomy, model capability, or benchmark performance. · Read the membership rule.

How it works under the hood

Seven mechanisms mapped from first-party records. These labels describe what the harness provides, not how intelligent its model is.

7/7layers documented
  • Execution & isolationWorkspace isolationDocumented mechanism, not a performance score.
  • Tooling & integrationsExtensible + browserDocumented mechanism, not a performance score.
  • Context & statePersistent stateDocumented mechanism, not a performance score.
  • Lifecycle & recoverySession resumeDocumented mechanism, not a performance score.
  • ObservabilityLogs/transcriptsDocumented mechanism, not a performance score.
  • VerificationTool-assistedDocumented mechanism, not a performance score.
  • Governance & permissionsPolicy controlsDocumented mechanism, not a performance score.

Public code audit

5/5public artifacts present
Security policy
Present at inspected commit
CI workflow
Present at inspected commit
Automated tests
Present at inspected commit
Evaluation assets
Present at inspected commit
Contributor documentation
Present at inspected commit

The 17.1.7 monorepo has broad automated tests, one public workflow, evaluation kernels, context-compression evaluation assets, and local-model engineering experiments. These project-owned artifacts and README performance claims lack the complete independent protocol required for a product benchmark.

Inspect commit d16c6168c86f40fc44f25118c2fd06fe160fcb93, checked 2026-07-28

Measured configurations

No benchmark run passes the full metadata admission policy for this harness yet. Missing data is not scored as zero.

Benchmark policy and all runs

Capability support

Documented first-class product support, checked against the sources below.

External tools (MCP)
DocumentedProduct-supported MCP integrationSource · checked 2026-07-28The source establishes the mechanism, not its quality or availability in every mode.
Local models
DocumentedLocal or self-hosted model pathSource · checked 2026-07-28The source establishes the mechanism, not its quality or availability in every mode.
Agent parallelism
DocumentedDelegated or parallel agent workflowSource · checked 2026-07-28The source establishes the mechanism, not its quality or availability in every mode.
Runs without an open UI
DocumentedNon-interactive or automation surfaceSource · checked 2026-07-28The source establishes the mechanism, not its quality or availability in every mode.
Browser control
DocumentedBuilt-in or product-supported browser controlSource · checked 2026-07-28The source establishes the mechanism, not its quality or availability in every mode.
Isolated execution
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Undo file changes
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.

Primary evidence

Each capability claim is tied to a first-party record and a verification date.

Product record checked 2026-07-28
Oh My Pi repositoryPinned 17.1.7 source, LSP, debugger, browser, subagents, tests, and MIT licenseOfficial repositoryOh My Pi 17.1.7 releaseCurrent stable package release, tool-input rewriting, schema revalidation, approval routing, and distribution version verificationOfficial announcementProduct overviewInstallation, terminal entry point, provider breadth, tools, agents, and extension surfaceOfficial docsApproval modesDefault yolo posture, tiers, per-tool policy, destructive overrides, ACP, and subagent behaviorOfficial docsSettings referenceApproval defaults, feature gates, project settings, memory, MCP, and task isolationOfficial docsProvider referenceCloud subscriptions, API providers, local runtimes, model roles, and fallback configurationOfficial docsRPC protocol referenceNDJSON automation, host tools, UI requests, and subagent event streamsOfficial docsTask subagentsParallel task batches, typed results, background execution, agent discovery, and workspace isolationOfficial docs
View 10 additional sources
Additional first-party evidence10 sources
Browser toolPuppeteer and CDP control, Chromium download, applications, screenshots, network, and process effectsOfficial docsMCP configurationLocal and remote transports, OAuth, profiles, discovery, credentials, timeouts, and enablementOfficial docsCheckpoint toolDisabled-by-default conversation marker, in-memory scope, persistence, and absent file snapshotOfficial docsRewind toolConversation pruning, retained report, branch persistence, and absent filesystem recoveryOfficial docsAutonomous memoryOptional persistent project memory, extraction, consolidation, redaction, scope, and defaultsOfficial docsACP implementationAgent Client Protocol entry point, editor integration, runtime options, and headless surfaceOfficial repositoryEvaluation runtimeProject-owned execution kernels and engineering tests rather than independent coding outcomesOfficial repositoryRuntime hooksCurrent extension-backed hook path, lifecycle events, pre-tool blocking or input rewriting, fail-closed interception, and loading behaviorOfficial docsRuntime extensionsIn-process extension runtime, tools, handlers, tool-input rewriting with schema and approval rechecks, session APIs, timers, and absent isolationOfficial docsSecret obfuscationDisabled-by-default outbound secret masking, environment and file discovery, reversible placeholders, and configurationOfficial docs