@hiai-gg/hiai-opencode 0.1.5 → 0.1.7

package/.env.example

@@ -11,14 +11,7 @@
 # - anthropic/claude-3-5-sonnet-latest
 #
 # Do not put model provider keys here for normal OpenCode usage.
-# Optional fallback only for headless/non-Connect workflows:
-# OPENROUTER_API_KEY=your_openrouter_api_key_here
-# OPENAI_API_KEY=your_openai_api_key_here
-# ANTHROPIC_API_KEY=your_anthropic_api_key_here
-# DEEPSEEK_API_KEY=your_deepseek_api_key_here
-# GLM_API_KEY=your_glm_api_key_here
-# MINIMAX_API_KEY=your_minimax_api_key_here
-# QWEN_API_KEY=your_qwen_api_key_here
+# Connect model providers in OpenCode itself.
 
 # MCP Service Keys
 
@@ -31,6 +24,13 @@
 # Optional - Context7 code search MCP
 # CONTEXT7_API_KEY=your_context7_api_key_here
 
+# Optional - Exa remote websearch MCP. Recommended for higher limits.
+# EXA_API_KEY=your_exa_api_key_here
+
+# Optional - Tavily remote websearch MCP.
+# Required only when hiai-opencode.json uses provider "tavily".
+# TAVILY_API_KEY=your_tavily_api_key_here
+
 # Local Model Helpers
 
 # Optional - Ollama local model endpoint
@@ -58,6 +58,19 @@
 # Optional - if Linux system browser deps cannot be installed, edit hiai-opencode.json
 # and add Playwright MCP args such as "--browser", "chrome" to use an existing browser.
 
+# Static MCP export
+
+# Optional - controls automatic .mcp.json export when the plugin loads.
+# Values: if-missing (default), always (managed files only), force, 0/false/no/off.
+# HIAI_OPENCODE_AUTO_EXPORT_MCP=if-missing
+
+# Optional - override static MCP export path.
+# HIAI_OPENCODE_MCP_EXPORT_PATH=./.mcp.json
+
+# Optional - CLI export safety mode.
+# Values: safe (default), force
+# HIAI_OPENCODE_EXPORT_MCP_MODE=safe
+
 # RAG MCP
 
 # Optional - RAG search tool endpoint

package/AGENTS.md

@@ -66,9 +66,14 @@ Manual alternative:
 
 ```bash
 opencode debug config
+hiai-opencode doctor
+hiai-opencode mcp-status
+hiai-opencode export-mcp .mcp.json
 opencode mcp list --print-logs --log-level INFO
 ```
 
+`opencode mcp list` reads static `.mcp.json` files in many OpenCode versions. Runtime MCP servers launched by the plugin may work but not appear there until a static config exists. The plugin auto-exports `.mcp.json` on startup when missing. In `always` mode it overwrites only hiai-managed exports unless forced. `hiai-opencode export-mcp .mcp.json` can refresh it manually.
+
 ## Post-Install Bootstrap Instructions
 
 When a user asks OpenCode or another agent to finish installing this plugin, follow this procedure.
@@ -110,13 +115,17 @@ When a user asks OpenCode or another agent to finish installing this plugin, fol
    - `FIRECRAWL_API_KEY`
    - `STITCH_AI_API_KEY`
    - `CONTEXT7_API_KEY`
-   - `GOOGLE_SEARCH_API_KEY`
+   - `EXA_API_KEY`
+   - `TAVILY_API_KEY`
    - `MEMPALACE_PYTHON`
    - `OPENCODE_RAG_URL`
    - `HIAI_PLAYWRIGHT_INSTALL_BROWSERS`
    - `HIAI_MCP_AUTO_INSTALL`
 7. Verify with:
    - `opencode debug config`
+   - `hiai-opencode doctor`
+   - `hiai-opencode mcp-status`
+   - `hiai-opencode export-mcp .mcp.json` when static MCP visibility is needed
    - `opencode mcp list --print-logs --log-level INFO`
 
 ### MCP Setup Matrix
@@ -126,14 +135,17 @@ When a user asks OpenCode or another agent to finish installing this plugin, fol
 | `playwright` | Node and npx are available | Helper launcher runs `@playwright/mcp@latest`; set `HIAI_PLAYWRIGHT_INSTALL_BROWSERS=1` to install Chromium on first start |
 | `sequential-thinking` | Node and npx are available | Helper launcher runs `@modelcontextprotocol/server-sequential-thinking` |
 | `firecrawl` | `FIRECRAWL_API_KEY` is set | Helper launcher runs `firecrawl-mcp` |
-| `mempalace` | `uv` is available, or Python 3.9+ with pip is available | Launcher prefers `uv`; otherwise uses Python and can run `python -m pip install --user mempalace` when `HIAI_MCP_AUTO_INSTALL` is not disabled |
+| `mempalace` | `uv` is available, or Python 3.9+ with pip is available | Launcher prefers `uv`; otherwise uses Python and can run `python -m pip install --user mempalace` when `HIAI_MCP_AUTO_INSTALL` is not disabled. Interpreter can be pinned via `mcp.mempalace.pythonPath` or `MEMPALACE_PYTHON` |
 | `rag` | User has a local or remote RAG endpoint | Uses `OPENCODE_RAG_URL`, defaulting to `http://localhost:9002/tools/search` |
 | `stitch` | `STITCH_AI_API_KEY` is set | Remote MCP endpoint |
 | `context7` | User wants Context7 docs/search | Remote MCP endpoint; use `CONTEXT7_API_KEY` if available |
+| `websearch` | User wants remote web search | Defaults to Exa; `EXA_API_KEY` optional for Exa, `TAVILY_API_KEY` required for Tavily |
+| `grep_app` | User wants GitHub/code search | Remote MCP endpoint; no key required |
 
 Playwright troubleshooting rules:
 
-- If `skill_mcp` says `MCP server "playwright" not found`, first load the `playwright` skill and check `hiai-opencode mcp-status`; do not report this as a browser dependency failure.
+- `skill_mcp` can call enabled hiai-opencode builtin MCPs and skill-embedded MCPs. If it reports a builtin MCP such as `firecrawl` as missing, check that the installed plugin version includes builtin fallback support and that `hiai-opencode.json` has that MCP enabled.
+- If `skill_mcp` says `MCP server "playwright" not found`, first check `hiai-opencode mcp-status`; do not report this as a browser dependency failure.
 - If Chromium reports missing Linux libraries (`libnspr4`, `libnss3`, `libatk-bridge`, `libgtk-3`, etc.), explain that MCP is present but the host lacks browser system dependencies.
 - Without sudo, try a system browser override in `hiai-opencode.json`, such as `--browser chrome` or `--browser msedge`.
 - If no browser path works, use `curl` only as degraded HTTP verification and explicitly say it is not a replacement for browser testing.
@@ -155,14 +167,16 @@ Check which services can run here:
 - playwright: node/npx; optionally HIAI_PLAYWRIGHT_INSTALL_BROWSERS=1.
 - sequential-thinking: node/npx.
 - firecrawl: FIRECRAWL_API_KEY.
-- mempalace: uv or Python 3.9+ with pip; MEMPALACE_PYTHON if needed; HIAI_MCP_AUTO_INSTALL controls first-run pip install.
+- mempalace: uv or Python 3.9+ with pip; set `mcp.mempalace.pythonPath` (or `MEMPALACE_PYTHON`) if needed; `HIAI_MCP_AUTO_INSTALL` controls first-run pip install.
 - rag: OPENCODE_RAG_URL or http://localhost:9002/tools/search.
 - stitch: STITCH_AI_API_KEY.
 - context7: optional CONTEXT7_API_KEY.
+- websearch: optional EXA_API_KEY for Exa, or TAVILY_API_KEY when provider is tavily.
+- grep_app: no key required.
 
 Report missing keys without printing secret values. Never invent or hardcode API keys.
 
-Run opencode debug config and opencode mcp list --print-logs --log-level INFO if available. If something is missing, propose or run only user-level/project-local install commands.
+Run hiai-opencode mcp-status and opencode debug config. If the user wants opencode mcp list visibility, run hiai-opencode export-mcp .mcp.json before opencode mcp list --print-logs --log-level INFO. If something is missing, propose or run only user-level/project-local install commands.
 ```
 
 ## Expected Agent State
@@ -216,6 +230,7 @@ The runtime loader is:
 Users configure only the 10 primary agent model slots under `models`: `bob`, `coder`, `strategist`, `guard`, `critic`, `designer`, `researcher`, `manager`, `brainstormer`, and `vision`.
 Hidden agents and task categories are derived internally in `src/config/defaults.ts`.
 Use fully qualified model IDs. Do not introduce local aliases like `hiai-fast`, `sonnet`, `fast`, or `high`.
+When helping a user choose model IDs, tell them to connect providers in OpenCode, run `opencode models`, and copy the exact `provider/model-id` strings into `hiai-opencode.json`. Do not invent provider prefixes.
 
 ## Change Map
 
@@ -345,6 +360,41 @@ Current MCP set:
 - `websearch`
 - `grep_app`
 
+## Writing And Website Copy
+
+Public-facing website/product copy is owned by `brainstormer`.
+`writer`, `copywriter`, `content-writer`, and `website-writer` are aliases for `brainstormer`.
+
+Use the `website-copywriting` skill for:
+
+- landing pages
+- hero sections
+- CTA labels
+- feature copy
+- positioning and naming
+- onboarding, empty states, and product microcopy
+
+Preferred invocation:
+
+```text
+task(subagent_type="brainstormer", load_skills=["website-copywriting"], run_in_background=false, description="Write landing page copy", prompt="...")
+```
+
+Use `designer` or `visual-engineering` for visual direction. Use `brainstormer` for words.
+
+## Manager Memory Stewardship
+
+Use `platform-manager` / `manager` when durable project memory or handoff state must stay current.
+
+Manager owns:
+
+- MemPalace decision hygiene: search first, deduplicate, write only durable decisions and important preferences.
+- Architecture memory updates: when architecture changes, update MemPalace and sync to RAG only if the configured RAG endpoint exposes write/upsert capability.
+- TODO hygiene: mark completed items complete, preserve unfinished tasks with blocker and next action, remove duplicate stale TODOs.
+- Session continuity: write concise handoff ledgers, not raw transcripts.
+
+Do not ask Manager to implement code. Use Coder for implementation and Manager for memory/task state.
+
 ## Skill Discovery Rules
 
 Default behavior is intentionally deterministic.
@@ -397,7 +447,8 @@ Common service keys:
 - `STITCH_AI_API_KEY`
 - `FIRECRAWL_API_KEY`
 - `CONTEXT7_API_KEY`
-- `GOOGLE_SEARCH_API_KEY`
+- `EXA_API_KEY`
+- `TAVILY_API_KEY`
 - `OLLAMA_BASE_URL`
 - `OLLAMA_MODEL`
 - `MEMPALACE_PYTHON`
@@ -405,6 +456,9 @@ Common service keys:
 - `OPENCODE_RAG_URL`
 - `HIAI_PLAYWRIGHT_INSTALL_BROWSERS`
 - `HIAI_MCP_AUTO_INSTALL`
+- `HIAI_OPENCODE_AUTO_EXPORT_MCP`
+- `HIAI_OPENCODE_MCP_EXPORT_PATH`
+- `HIAI_OPENCODE_EXPORT_MCP_MODE`
 
 ## Known Runtime Caveats
 

package/ARCHITECTURE.md

@@ -89,10 +89,13 @@ Prompting is not a single file and not a single directory. `src/agents` is the m
 These files create the top-level agent config objects:
 
 - Bob: [src/agents/bob.ts](src/agents/bob.ts)
-- Coder: `src/agents/coder/agent.ts`
+- Coder: [src/agents/coder/agent.ts](src/agents/coder/agent.ts)
 - Strategist: `src/agents/strategist/*`
-- Guard: `src/agents/guard/agent.ts`
-- Critic: `src/agents/critic/agent.ts`
+- Guard: [src/agents/guard/agent.ts](src/agents/guard/agent.ts)
+- Critic: [src/agents/critic/agent.ts](src/agents/critic/agent.ts)
+- Designer: [src/agents/designer.ts](src/agents/designer.ts)
+- Brainstormer: [src/agents/brainstormer.ts](src/agents/brainstormer.ts)
+- Agent Skills: [src/agents/agent-skills.ts](src/agents/agent-skills.ts)
 - Vision: [src/agents/ui.ts](src/agents/ui.ts)
 - Manager: [src/agents/platform-manager.ts](src/agents/platform-manager.ts)
 - Researcher: [src/agents/researcher.ts](src/agents/researcher.ts)

package/LICENSE.md

@@ -31,7 +31,6 @@ Some are:
 | Sequential Thinking | [modelcontextprotocol/servers](https://github.com/modelcontextprotocol/servers) | external MCP |
 | Firecrawl MCP | [firecrawl-ai/firecrawl-mcp-server](https://github.com/firecrawl-ai/firecrawl-mcp-server) | external MCP |
 | Context7 MCP | [upstash/context7-mcp](https://github.com/upstash/context7-mcp) | external MCP |
-| Websearch cited | [ghoulr/opencode-websearch-cited](https://github.com/ghoulr/opencode-websearch-cited) | search integration influence |
 | bun-pty / PTY ecosystem | [shekohex/opencode-pty](https://github.com/shekohex/opencode-pty) | PTY/runtime integration influence |
 
 ## What Is Bundled vs External

package/README.md

@@ -1,5 +1,7 @@
 # hiai-opencode
 
+[![CI](https://github.com/HiAi-gg/hiai-opencode/actions/workflows/ci.yml/badge.svg)](https://github.com/HiAi-gg/hiai-opencode/actions/workflows/ci.yml)
+
 `hiai-opencode` is an OpenCode plugin that wires together:
 
 - a curated multi-agent runtime
@@ -31,21 +33,47 @@ Keep OpenCode plugins separate from MCP servers. Do not add MCP server packages
 
 Check which MCP services can run on this machine, update hiai-opencode.json, install only missing user-level or project-local dependencies, and report missing API keys without printing secret values.
 
-Then run opencode debug config and opencode mcp list --print-logs --log-level INFO if available.
+Then run `hiai-opencode doctor`, `hiai-opencode mcp-status`, and `opencode debug config`.
 ```
 
 For the full operator playbook, see [AGENTS.md](AGENTS.md). 🤖
 
+## Agents
+
+| Agent | Role | When to use |
+|-------|------|-------------|
+| `Bob` | Orchestrator, router, distributor | Entry point; complex tasks needing multi-agent coordination |
+| `Coder` | Deep implementation, focused execution | Complex features, refactors, deep work |
+| `Sub` | Bounded cheap executor | Small targeted changes, quick fixes |
+| `Strategist` | Planning, architecture, pre-check | Scope definition, architectural decisions |
+| `Guard` | Final acceptor, workflow enforcer | Closure validation, output acceptance |
+| `Critic` | Review gate, high-accuracy verification | Plan review, code review, regression catch |
+| `Researcher` | Local + external search | Codebase exploration, documentation discovery |
+| `Designer` | UI/visual, creative direction | Visual problems, UX decisions, branding |
+| `Brainstormer` | Ideation, content, copy | Landing pages, CTA, feature copy, onboarding |
+| `Vision` | Image/PDF/layout analysis | Visual inspection, multimodal interpretation |
+| `Manager` | Memory, bootstrap, ledger | Durable state, session continuity, project init |
+
+## Modes (Task Routing)
+
+Mode determines prompt append, variant, and reasoning effort. The executor agent is selected via `mode → agent` mapping.
+
+| Mode | Agent | Prompt variant | When to use |
+|------|-------|----------------|-------------|
+| `quick` | `sub` | Fast bounded | Small targeted changes |
+| `writing` | `brainstormer` | Docs/prose | Content, i18n, copy |
+| `deep` | `coder` | Deep reasoning | Complex implementation |
+| `ultrabrain` | `strategist` | Plan-only | Architecture, hard logic |
+| `visual-engineering` | `designer` | UI/visual | Visual problems |
+| `artistry` | `designer` | Creative | Brand, SEO, creative |
+| `git` | `platform-manager` | Git ops | Version control operations |
+| `bounded` | `sub` | Mid-tier bounded | Moderate effort changes |
+| `cross-module` | `coder` | Deep substantial | Multi-component changes |
+
 ## What You Get
 
-- Visible primary agents: `Bob`, `Coder`, `Strategist`, `Guard`, `Critic`, `Designer`, `Researcher`, `Manager`, `Brainstormer`, `Vision`
-- Hidden system or compatibility agents: `Agent Skills`, `Sub`, `build`, `plan`
-- Task routing model:
-  - category-based execution routes through `Coder`
-  - `quick`, `writing`, and `unspecified-low` are the fast bounded Coder contour
-  - `deep`, `ultrabrain`, `visual-engineering`, `artistry`, and `unspecified-high` are the deep Coder contour
-  - `Critic` and `Researcher` are selected explicitly
-  - `Designer`, `Brainstormer`, `Manager`, and `Vision` are direct callable specialists
+- **10 visible primary agents** + **4 hidden system agents** (Agent Skills, Sub, build, plan)
+- **Mode-based task routing** via `task(category=..., ...)` or `task(mode=..., ...)`
 - Skill materialization into OpenCode's `skills/` view
 - MCP wiring for `playwright`, `stitch`, `sequential-thinking`, `firecrawl`, `rag`, `mempalace`, `context7`, plus remote `websearch` and `grep_app`
 - LSP wiring for TypeScript, Svelte, Python, Bash, and ESLint
@@ -67,6 +95,8 @@ Optional, depending on which services you want:
 - `FIRECRAWL_API_KEY` for Firecrawl
 - `STITCH_AI_API_KEY` for Stitch
 - `CONTEXT7_API_KEY` for Context7
+- `EXA_API_KEY` for higher Exa websearch limits
+- `TAVILY_API_KEY` when `mcp.websearch.provider` is `tavily`
 - Python 3.9+ or `uv` for MemPalace
 - a running RAG endpoint if you enable `rag`
 - local language servers if you want LSP beyond the npm-bootstrapped helpers
@@ -121,22 +151,22 @@ If you installed only from npm/OpenCode and do not have this repository checked
 ```json
 {
   "models": {
-    "bob": { "model": "openrouter/anthropic/claude-3.5-opus", "recommended": "xhigh" },
-    "coder": { "model": "openrouter/anthropic/claude-3.5-sonnet", "recommended": "high" },
-    "strategist": { "model": "openrouter/z-ai/glm-5.1", "recommended": "high" },
-    "guard": { "model": "openrouter/openai/gpt-4o", "recommended": "middle" },
-    "critic": { "model": "openrouter/qwen/qwen2.5-72b-instruct", "recommended": "high" },
-    "designer": { "model": "openrouter/google/gemini-3.1-pro", "recommended": "design" },
-    "researcher": { "model": "openrouter/google/gemini-2.0-flash", "recommended": "fast" },
-    "manager": { "model": "openrouter/google/gemini-2.0-flash", "recommended": "fast" },
-    "brainstormer": { "model": "openrouter/kimi/kimi-latest", "recommended": "writing" },
-    "vision": { "model": "openrouter/google/gemini-3.1-pro", "recommended": "vision" }
+    "bob": { "model": "openrouter/moonshotai/kimi-k2.6", "recommended": "xhigh" },
+    "coder": { "model": "openrouter/minimax/minimax-m2.7", "recommended": "high" },
+    "strategist": { "model": "openrouter/anthropic/claude-opus-latest", "recommended": "high" },
+    "guard": { "model": "openrouter/qwen/qwen3.6-plus", "recommended": "middle" },
+    "critic": { "model": "openrouter/xiaomi/mimo-v2.5-pro", "recommended": "high" },
+    "designer": { "model": "openrouter/google/gemini-3.1-pro-preview", "recommended": "design" },
+    "researcher": { "model": "openrouter/deepseek/deepseek-v4-flash", "recommended": "fast" },
+    "manager": { "model": "openrouter/qwen/qwen3.5-9b", "recommended": "fast" },
+    "brainstormer": { "model": "openrouter/mistralai/mistral-small-2603", "recommended": "writing" },
+    "vision": { "model": "openrouter/google/gemma-4-26b-a4b-it", "recommended": "vision" }
   },
   "mcp": {
     "playwright": { "enabled": true },
     "sequential-thinking": { "enabled": true },
     "firecrawl": { "enabled": true },
-    "mempalace": { "enabled": true },
+    "mempalace": { "enabled": true, "pythonPath": "{env:MEMPALACE_PYTHON:-./.venv/bin/python}" },
     "rag": { "enabled": false },
     "stitch": { "enabled": false },
     "context7": { "enabled": true }
@@ -153,10 +183,19 @@ This plugin only reads the 10 model IDs in `models`. Internal routing derives hi
 
 Use OpenCode Connect to authorize the providers behind your configured model IDs. Then add only the service keys for MCP or search integrations you actually use:
 
+```bash
+opencode models
+```
+
+Use the exact model IDs printed by OpenCode in `hiai-opencode.json`. For example, `openrouter/minimax/minimax-m2.7` routes through OpenRouter, while `minimax/minimax-m2.7` routes through a direct Minimax provider only if that provider is connected in OpenCode.
+
 ```bash
 export FIRECRAWL_API_KEY=...
 export STITCH_AI_API_KEY=...
 export CONTEXT7_API_KEY=...
+export EXA_API_KEY=...
+# or, if mcp.websearch.provider is "tavily":
+export TAVILY_API_KEY=...
 ```
 
 See [Environment Variables And Keys](#environment-variables-and-keys) for the full list.
@@ -165,15 +204,19 @@ See [Environment Variables And Keys](#environment-variables-and-keys) for the fu
 
 ```bash
 opencode
+hiai-opencode doctor
 hiai-opencode mcp-status
+hiai-opencode export-mcp .mcp.json
 opencode debug config
 opencode mcp list --print-logs --log-level INFO
 ```
 
-`opencode mcp list` may not show plugin-provided MCP servers in every OpenCode version. If that happens, use `opencode debug config` and startup logs as the source of truth.
+`opencode mcp list` reads static `.mcp.json` files in many OpenCode versions. Runtime MCP servers launched by plugins may work but not appear there. If you want `opencode mcp list` visibility, run `hiai-opencode export-mcp .mcp.json` first.
 
 `hiai-opencode mcp-status` is the fastest visibility check. It does not change OpenCode config; it reports config location, enabled MCP services, missing keys, and basic local runtime availability.
 
+`hiai-opencode doctor` is the broader install/runtime diagnostic. It includes MCP status, static `.mcp.json` freshness, OpenCode Connect visibility, skill materialization, agent naming/count checks, LSP runtime checks, MemPalace Python source selection, and real MCP tool probes.
+
 ## Development Install
 
 Direct npm install is only needed for development or inspection:
@@ -210,7 +253,7 @@ Enable only services that can run on this machine:
 - playwright: requires node/npx; optionally set HIAI_PLAYWRIGHT_INSTALL_BROWSERS=1 before first run if browser binaries are needed.
 - sequential-thinking: requires node/npx.
 - firecrawl: requires FIRECRAWL_API_KEY.
-- mempalace: requires uv or Python 3.9+ with pip; set MEMPALACE_PYTHON if needed. Leave HIAI_MCP_AUTO_INSTALL enabled unless the user forbids package installation.
+- mempalace: requires uv or Python 3.9+ with pip; set `mcp.mempalace.pythonPath` (or `MEMPALACE_PYTHON`) if needed. Leave `HIAI_MCP_AUTO_INSTALL` enabled unless the user forbids package installation.
 - rag: requires OPENCODE_RAG_URL or a running local endpoint at http://localhost:9002/tools/search.
 - stitch: requires STITCH_AI_API_KEY.
 - context7: works without a key but use CONTEXT7_API_KEY if available.
@@ -219,6 +262,8 @@ Check .env.example, report missing keys without printing secret values, and neve
 
 Run verification commands where available:
 - opencode debug config
+- hiai-opencode mcp-status
+- hiai-opencode export-mcp .mcp.json
 - opencode mcp list --print-logs --log-level INFO
 
 If a dependency is missing, install only user-level or project-local dependencies, explain every command before running it, and do not use sudo/admin rights unless the user explicitly asks.
@@ -239,6 +284,7 @@ Runtime loader for the bundled canonical config:
 If you want to change model selection, edit the 10 entries in `models`. Do not add category-specific model choices unless you are intentionally developing the plugin internals.
 
 Use fully qualified model IDs. Do not introduce local aliases like `hiai-fast`, `sonnet`, `fast`, or `high`.
+After connecting providers in OpenCode, run `opencode models` and copy the exact model IDs from that output into `hiai-opencode.json`.
 
 ### Prompting
 
@@ -257,6 +303,7 @@ Important prompt entrypoints:
 - `Vision`: [src/agents/ui.ts](src/agents/ui.ts)
 - `Manager`: [src/agents/platform-manager.ts](src/agents/platform-manager.ts)
 - `Researcher`: [src/agents/researcher.ts](src/agents/researcher.ts)
+- `Brainstormer` / `Writer`: [src/agents/brainstormer.ts](src/agents/brainstormer.ts)
 
 Name mapping and visibility:
 
@@ -273,6 +320,22 @@ Skill materialization logic:
 
 - [src/features/builtin-skills/materialize.ts](src/features/builtin-skills/materialize.ts)
 
+Built-in helper skills include browser automation, frontend UI/UX, review, git workflow, hiai-opencode setup, AI slop cleanup, and `website-copywriting`.
+
+Website/product copy should use:
+
+```text
+task(subagent_type="brainstormer", load_skills=["website-copywriting"], ...)
+```
+
+`writer`, `copywriter`, and `content-writer` are aliases for `brainstormer`.
+
+Manager memory stewardship:
+
+- Use `task(subagent_type="platform-manager", ...)` or `task(subagent_type="manager", ...)` for MemPalace cleanup, session ledgers, TODO hygiene, and architecture decision handoff.
+- Manager writes only durable decisions and important project state. It should not dump raw chat logs into memory.
+- RAG is retrieval-first by default; Manager syncs architecture summaries to RAG only when the configured endpoint exposes write/upsert capability.
+
 Skill discovery defaults:
 
 ```json
@@ -321,7 +384,8 @@ Important service variables:
 - `STITCH_AI_API_KEY`
 - `FIRECRAWL_API_KEY`
 - `CONTEXT7_API_KEY`
-- `GOOGLE_SEARCH_API_KEY`
+- `EXA_API_KEY`
+- `TAVILY_API_KEY`
 - `OLLAMA_BASE_URL`
 - `OLLAMA_MODEL`
 - `MEMPALACE_PYTHON`
@@ -329,6 +393,8 @@ Important service variables:
 - `OPENCODE_RAG_URL`
 - `HIAI_PLAYWRIGHT_INSTALL_BROWSERS`
 - `HIAI_MCP_AUTO_INSTALL`
+- `HIAI_OPENCODE_AUTO_EXPORT_MCP`
+- `HIAI_OPENCODE_MCP_EXPORT_PATH`
 
 Optional headless or non-Connect fallback variables are documented in [.env.example](.env.example), but they are not required for normal OpenCode model auth.
 
@@ -342,7 +408,9 @@ The user-facing MCP switchboard is the `mcp` object in `hiai-opencode.json`:
 {
   "mcp": {
     "playwright": { "enabled": true },
-    "mempalace": { "enabled": false }
+    "mempalace": { "enabled": false },
+    "websearch": { "enabled": true, "provider": "exa" },
+    "grep_app": { "enabled": true }
   }
 }
 ```
@@ -353,7 +421,7 @@ The source of truth for default MCP wiring is `src/mcp/registry.ts`. Change that
 
 - `stitch`
 - `context7`
-- `websearch`
+- `websearch`: defaults to Exa remote MCP. `EXA_API_KEY` is optional for Exa; set `"provider": "tavily"` and `TAVILY_API_KEY` to use Tavily.
 - `grep_app`
 
 ### Works with local helper bootstrap
@@ -393,7 +461,7 @@ If sudo is not available:
 
 ### Needs upstream runtime or extra setup
 
-- `mempalace`: prefers `uv`; otherwise uses Python. If `HIAI_MCP_AUTO_INSTALL` is not `0`, `false`, or `no`, the launcher can run `python -m pip install --user mempalace` on first start.
+- `mempalace`: prefers `uv`; otherwise uses Python. You can force interpreter selection via `mcp.mempalace.pythonPath` or `MEMPALACE_PYTHON`. If `HIAI_MCP_AUTO_INSTALL` is not `0`, `false`, or `no`, the launcher can run `python -m pip install --user mempalace` on first start.
 - `rag`: requires your own running endpoint
 
 ### Important Windows note
@@ -415,12 +483,26 @@ The plugin now emits startup warnings for common misconfiguration, including:
 Available CLI:
 
 ```bash
+hiai-opencode doctor
 hiai-opencode mcp-status
+hiai-opencode export-mcp .mcp.json
+```
+
+By default, the plugin auto-exports `.mcp.json` at workspace startup when the file is missing. This closes the visibility gap where runtime plugin MCP works but `opencode mcp list` only reads static files. Control it with:
+
+```bash
+export HIAI_OPENCODE_AUTO_EXPORT_MCP=if-missing  # default
+export HIAI_OPENCODE_AUTO_EXPORT_MCP=always      # overwrite only managed hiai-opencode exports
+export HIAI_OPENCODE_AUTO_EXPORT_MCP=force       # force overwrite even non-managed files
+export HIAI_OPENCODE_AUTO_EXPORT_MCP=0           # disable auto-export
+export HIAI_OPENCODE_MCP_EXPORT_PATH=.mcp.json   # override path
+export HIAI_OPENCODE_EXPORT_MCP_MODE=safe        # export-mcp command mode: safe|force
 ```
 
 Inside OpenCode, use the slash command:
 
 ```text
+/doctor
 /mcp-status
 ```
 
@@ -435,14 +517,13 @@ MCP Servers:
 ⚠️  stitch              - enabled, API key missing (STITCH_AI_API_KEY)
 ```
 
-Planned follow-up commands:
+`hiai-opencode export-mcp` writes a standard `.mcp.json` so hosts whose `mcp list` ignores plugin runtime MCP can still show the same servers statically. Exports are marker-tagged as hiai-managed; by default, the command avoids overwriting non-managed files unless `HIAI_OPENCODE_EXPORT_MCP_MODE=force` is set.
 
-- `hiai-opencode doctor`: report config location, MCP status, missing keys, and local dependency checks.
-- `hiai-opencode export-mcp`: generate a standard `.mcp.json` for hosts that do not expose plugin-provided MCP servers through `opencode mcp list`.
-
-Until those commands ship, use:
+Use:
 
 ```bash
+hiai-opencode mcp-status
+hiai-opencode export-mcp .mcp.json
 opencode debug config
 opencode mcp list --print-logs --log-level INFO
 ```
@@ -464,7 +545,6 @@ opencode mcp list --print-logs --log-level INFO
 | Sequential Thinking | [modelcontextprotocol/servers](https://github.com/modelcontextprotocol/servers) | external MCP |
 | Firecrawl MCP | [firecrawl-ai/firecrawl-mcp-server](https://github.com/firecrawl-ai/firecrawl-mcp-server) | external MCP |
 | Context7 MCP | [upstash/context7-mcp](https://github.com/upstash/context7-mcp) | external MCP |
-| Websearch cited | [ghoulr/opencode-websearch-cited](https://github.com/ghoulr/opencode-websearch-cited) | search integration influence |
 | bun-pty / PTY ecosystem | [shekohex/opencode-pty](https://github.com/shekohex/opencode-pty) | PTY/runtime integration influence |
 
 ## Build And Publish
@@ -492,7 +572,7 @@ Before publishing:
 1. run `bun run build`
 2. run `npm pack --dry-run`
 3. verify `debug config`
-4. verify `mcp list`
+4. run `hiai-opencode export-mcp .mcp.json` if you need static `mcp list` visibility
 
 Publish: