WP HoloDev combines AI coding models with a controlled development harness built around real WordPress and software development. It gives agents project intelligence, development tools, specialist skills, persistent memory, testing, browser access and release workflows while keeping execution under harness control.
Inspect existing projects, make coordinated changes across multiple files, run development commands, bring in specialist agents, connect MCP tools, validate the result and move completed work through review, packaging, deployment and rollback from one local development environment.
You choose the AI models. WP HoloDev provides the development system around them.
Control How the AI Works
WP HoloDev gives you direct control over what kind of work the AI performs, which models it can use and how much agent delegation it is allowed to create.
Full Auto, Semi-Auto and Manual Model Assignment
Choose how much control WP HoloDev has over model selection for each project.
Full Auto selects the cheapest suitable model for the work. Semi-Auto chooses only from the exact models and provider connections you allow. Manual locks the entire job to one exact provider, account and model.
Full, Low and No Subagents
Agent Level controls how much delegation a development job can use independently of model selection.
Full allows specialist subagents when they are genuinely useful. Low limits the entire job to four subagents. None keeps everything with the main agent while still allowing tools, skills and validation.
Five Main Development Modes
Set the type of work the harness should optimise for with Coding, WordPress Coding, Design, WordPress Design or Cyber Security.
The active mode influences model suitability, prompts, skills, tools and validators without changing your model or agent-level controls.
Persistent Jobs That Continue Until the Work Is Done
WP HoloDev tracks development as structured jobs with requirements, TODOs, progress and verification rather than relying on a model simply saying that it has finished.
If actionable work remains, the harness can continue the same job instead of accepting a premature final response. Completed work, policy settings and outstanding tasks remain part of the stored job state.
Automatic Skills and Tools
You do not need to manually choose every capability before starting a task.
WP HoloDev can select relevant skills, native tools and validators from the project, task and active Main Mode while keeping execution inside the same permission and security system.
Independent Project Policies
Each project can keep its own Main Mode, Model Assignment policy, Semi-Auto allowlist, Manual route and Agent Level.
Running jobs snapshot those settings when they start, so changing a project default later does not silently change work already in progress.
Use the Right Model for Each Job
Connect the AI services you want and let WP HoloDev use them according to rules you control rather than tying the entire development environment to one model or provider.
Cheapest Suitable Model Routing
Full Auto does not simply select the cheapest model or the strongest model.
WP HoloDev evaluates the capabilities, context requirements, tool requirements, availability and cost of the job, then selects the lowest-cost exact route that is sufficient for the work.
Exact Manual Model Locking
Manual means manual. Every model-backed part of the job uses the exact provider connection and model you selected.
This includes the main agent, planning, review, recovery, continuation and permitted subagents. WP HoloDev does not silently swap to another account or model.
Controlled Semi-Auto Routing
Semi-Auto lets WP HoloDev choose intelligently while keeping model selection inside a project-specific allowlist that you define.
Different accounts using the same provider or model remain separate exact routes, so the harness cannot silently move between connections.
Multiple AI Providers
WP HoloDev supports multiple cloud and local model providers, including OpenAI, Anthropic, OpenRouter, OpenCode, LM Studio, WP HoloDev AI and compatible custom endpoints.
You can combine different providers in one development environment without merging their credentials, model catalogues or project policies.
Multiple Accounts on the Same Provider
Two connections to the same provider remain separate throughout discovery, routing, project settings and diagnostics.
This means an OpenAI model on one account is not treated as interchangeable with the same model name on another account.
Local Models with LM Studio
Connect an LM Studio server and use compatible models running on your own computer through the same WP HoloDev workflow as cloud models.
Local models can participate in normal model selection without requiring a separate development environment.
Live Connection-Specific Model Discovery
Supported providers can expose their available models directly to WP HoloDev.
Model catalogues are maintained per connection rather than being pooled across accounts, with capability and cost metadata used where available.
Usage and Cost Tracking
Token usage and estimated model cost can be recorded against development jobs so you can see what AI-assisted development is consuming.
Jobs use unlimited cumulative usage by default, while explicit token or cost budgets can still be applied when you want a hard limit.
A Development Harness, Not Just a Chat Window
WP HoloDev gives AI models a structured development environment with project intelligence, controlled tools, persistent state and specialist capabilities.
Harness-Controlled Tool Execution
Agents can inspect files, search projects, edit code, run development commands, use WordPress tools, perform validation and interact with supported external capabilities.
Every native tool invocation passes through WP HoloDev's ActionRegistry, argument validation, permissions and project boundaries rather than giving the model unrestricted computer access.
Repository Intelligence
WP HoloDev analyses project structure and code so agents can identify files, symbols and relationships relevant to the task.
For WordPress projects this includes specialist understanding of plugin and theme structures, hooks, REST endpoints, database access and other WordPress development patterns.
Specialist Subagents
Larger tasks can be divided between specialist agents for focused investigation, implementation, testing, debugging, design, WordPress work or defensive security review.
Subagents inherit the running job's model, mode, security and project policies instead of operating as unrestricted independent agents.
Persistent Project Memory
Useful knowledge can remain associated with a project between development sessions instead of disappearing when one conversation ends.
Project information, decisions and remembered context remain inspectable so persistent knowledge can support future work without becoming an invisible source of truth.
Detailed Execution Trace
Your normal conversation can stay focused on the development objective while WP HoloDev separately records operational work carried out behind it.
Inspect agents, actual model routes, tool calls, validation, lifecycle decisions and other execution details when you need to understand how a task was performed.
Direct Development Commands
Common harness operations are available through slash commands such as /check, /test, /review, /build, /deploy, /rollback, /models, /skills, /tools, /memory and /status.
You can still work naturally through conversation. Commands provide direct access to common workflows when you want them.
Specialist Skills and Model-Aware Prompts
WP HoloDev combines reusable development skills with prompts compiled for the actual task, project, working mode and model being used.
24 Bundled Development Skills
WP HoloDev ships with 24 bundled skills covering general software engineering and specialist WordPress development.
They include code review, debugging, test-driven development, implementation, architecture, WordPress plugin development, performance, PHPStan, REST API development, WP-CLI operations, WordPress.org guidelines and Holographic Plugin Check.
Automatic Skill Selection
The harness can choose relevant skills automatically from the task, Main Mode and project context.
Skills are added when useful rather than loading every specialist instruction into every model request.
Model-Specific Prompt Profiles
Different models can receive instructions tailored to the way they work rather than every provider being given one giant generic prompt.
If Full Auto or Semi-Auto reroutes a job to another model, WP HoloDev recompiles the prompt for the model that actually receives the work.
Extensible Skill System
Bundled skills are not the limit of the system. WP HoloDev can discover additional user and project skills while retaining source and trust information.
Skill integrity can be checked with content checksums, while project-supplied content does not automatically receive the same trust as bundled capabilities.
Real Validation Before the AI Says It Is Finished
WP HoloDev backs model reasoning with actual development tools, structured requirements and deterministic validation.
Deterministic Validation
WP HoloDev can run real project validators including PHP syntax checks, WordPress Coding Standards, PHPStan, PHPUnit, Plugin Check and project-specific test or build commands where available.
The results become part of the job evidence rather than relying on the model to judge its own output.
Structured Completion Requirements
Outstanding TODOs, required work and active delegated tasks are tracked separately from the model's final prose.
A model saying “done” does not make unfinished structured work disappear. Completion is decided by the harness using the state of the actual job.
Independent Code Review
A development job can include a separate review stage so the agent that wrote the change is not the only system evaluating its own work.
Review can use project changes, test evidence and findings rather than relying solely on an implementation summary.
Regression-Aware Validation
Existing project findings can be recorded before changes are made so later validation can distinguish older technical debt from regressions introduced by the current work.
This gives the agent clearer evidence about whether its own changes made the project better or worse.
Browser Testing with Playwright
WP HoloDev includes Playwright-based browser automation so agents can inspect the behaviour produced by their code rather than stopping at static analysis.
Pages can be loaded, interfaces exercised and screenshots captured as evidence during testing and debugging.
Isolated WordPress Test Sites
When Docker Desktop is available, WP HoloDev can provision isolated WordPress environments for runtime testing away from real customer or production sites.
Docker is optional and is not required simply to install or use the harness.
Holographic Plugin Check
The bundled Holographic Plugin Check workflow provides focused final QA for WordPress plugins, including code quality, security, optimisation, release readiness and WordPress.org considerations.
It can also be accessed quickly through /check and its /holo and /audit aliases.
Source-Bound Validation Evidence
Final validation can be tied to the exact source that passed the checks.
If the source changes afterwards, previous evidence can be invalidated rather than being treated as proof for code that was never actually tested.
Security Around the AI, Not Inside It
The model is allowed to be useful without being trusted as the security boundary for your development environment.
Project Filesystem Boundaries
Project file operations are constrained to the intended workspace using canonical path checks and protections against traversal, Windows reparse points, junctions and filesystem-link escapes.
This prevents an ordinary development request from becoming arbitrary filesystem access because the model generated a dangerous path.
ActionRegistry Security
Native tools are registered, validated and executed through one controlled action layer.
Automatic tools, explicit tool commands and subagent tools do not get separate routes around the same permission, validation and project-isolation checks.
Protected Provider Credentials
Provider credentials are protected at rest and kept separate from ordinary project instructions and model prompts.
Recursive redaction and credential-shaped metadata checks help keep tokens and other secrets out of diagnostics, logs, stored metadata and release packages.
Untrusted Project Content Stays Untrusted
Source code, comments, READMEs and tool output are treated as project data rather than automatically becoming trusted harness instructions.
A malicious repository file cannot promote itself into system policy simply by telling the model to ignore WP HoloDev's controls.
Safer Network and Browser Access
Network and browser capabilities include protections around private or reserved targets, credential-bearing URLs, DNS rebinding and validated destination addresses.
These controls are designed to preserve legitimate local development services without turning arbitrary project input into unrestricted internal network access.
Safe ZIP and Archive Handling
Archive handling validates paths, destinations, encoded input, entry counts and extracted size before installation.
Staged extraction and additional path protections help defend against traversal, unsafe Windows filenames, extraction races and other hostile archive techniques.
Cross-Project Isolation
Project files, settings, jobs, provider routes and WordPress development state remain scoped to the project they belong to.
Isolation checks cover both normal Windows development and Docker-backed WordPress workflows.
Local Application Protection
WP HoloDev's local application server includes host, origin, session and application-token controls rather than assuming that anything reaching localhost should automatically be trusted.
Cross-origin protections help prevent an unrelated browser page from silently invoking privileged local development actions.
Cancellation That Stops Real Work
Cancelling a job prevents new tool execution, provider continuation and delegated work from silently restarting it through stale asynchronous results.
Cancellation belongs to the harness rather than being a suggestion to the model.
Structured Audit History
Development activity can be recorded in a structured audit trail including tool use, model routes, job transitions, security decisions and validation activity.
This gives you a reviewable record of what the system actually did instead of relying solely on the final conversation.
MCP and External Development Capabilities
WP HoloDev can extend its native development tools with compatible Model Context Protocol servers without treating external capabilities as inherently trusted.
MCP Server Connections
Connect compatible local STDIO and HTTP-based MCP servers to add development capabilities to WP HoloDev.
Workspace validation, connection controls and network protections keep MCP inside the same overall project security model.
MCP Tools
Tools discovered from MCP servers can become available alongside WP HoloDev's native tools.
External tool identity remains tied to its MCP server and does not silently replace a native HoloDev capability.
MCP Resources and Prompts
Compatible servers can expose resources and reusable prompts in addition to executable tools.
WP HoloDev can discover and retrieve those capabilities when supported by the connected server.
MCP Still Goes Through the Harness
MCP is an extension layer, not a route around WP HoloDev security.
External capabilities remain subject to project boundaries, tool controls and auditability rather than automatically becoming trusted because an MCP server supplied them.
Release, Deployment and Rollback
Validated WordPress work can move through release preparation and controlled deployment without giving the coding model unrestricted hosting credentials.
Release Packaging
WP HoloDev can prepare distributable project ZIP files after development and validation are complete.
Release artefacts can be fingerprinted with cryptographic checksums and scanned so the harness can verify the package used later in the workflow.
Source-Bound Final Testing
Release validation can record a fingerprint of the exact source that passed the checks.
If the project changes afterwards, previous final-test evidence no longer represents the modified source.
Controlled WordPress Bridge
WP HoloDev can pair a project with a WordPress installation through the HoloDev deployment bridge.
Authenticated requests, project authorisation, timestamps, integrity checks and replay protections allow controlled deployment without exposing ordinary hosting credentials to the model.
Post-Deployment Verification
Deployment does not have to end when a ZIP reaches WordPress.
WP HoloDev can verify the destination after installation so deployment evidence remains part of the development workflow.
Rollback Support
If a deployment needs to be reversed, WP HoloDev provides a controlled workflow for returning to a previous verified deployment.
Recovery remains part of the deployment system instead of requiring an agent to improvise changes directly on the server.
A Practical Local Development Workspace
WP HoloDev runs locally on Windows and brings projects, models, jobs, skills, tools, testing and WordPress workflows together without requiring a separate server stack just to get started.
Simple Windows Setup
WP HoloDev includes its own Node.js runtime, local application server and SQLite database.
You do not need to install Python, PostgreSQL, Redis or Docker simply to run the harness. Extract the release and launch WP HoloDev.
Docker Only When You Need It
Docker Desktop is optional and is used when you choose isolated WordPress test environments or Docker-based development checks.
Normal project management, conversations and AI-assisted development do not require Docker to remain running.
Project-Based Workspace
Keep development work organised around individual projects with their own conversations, jobs, memory, model policies, working modes, validation and deployment information.
This keeps unrelated development contexts separated while giving each project persistent state.
Dark, Light and Electric Modes
Choose between Dark, Light and Electric interface modes so the development workspace can match the way you prefer to work.
Appearance remains separate from development behaviour and permissions.
Custom Main Agent Persona
Change the name, personality, tone and conversational style of the main WP HoloDev agent.
Persona is deliberately separated from technical authority. It cannot alter project boundaries, tool permissions, model policy, Agent Level or validation requirements.
Erin Included by Default
Fresh installations start with Erin as the default WP HoloDev coding personality.
Keep Erin, change her style, replace her with your own persona or disable the persona layer entirely.
WordPress Specialist, Not WordPress-Only
WP HoloDev provides dedicated WordPress coding, design, validation and deployment capabilities, but the underlying agent and development systems can also work with other software projects.
Use the Main Mode that fits the work rather than being locked into one development stack.
More Than a Coding Model in a Different Window
WP HoloDev combines persistent development jobs, intelligent model routing, Main Modes, controlled subagents, project intelligence, specialist skills, native tools, MCP integrations, deterministic validation, browser testing, security controls and WordPress workflows inside one harness.
You choose how much control the harness has, which models and accounts it may use, how much delegation is allowed and what kind of work it should optimise for. WP HoloDev then gives those models a controlled environment in which to inspect the project, perform the work and verify the result.
The AI supplies the intelligence. WP HoloDev supplies the development harness.