chore: remove dead opencode docs MCP

AGENTS.md

@@ -35,7 +35,7 @@ For Hatch resources, use only `aws-staging`, `aws-prod`, `argocd-staging-eks`, `
 - Library and framework questions stay with Context7 even when phrased as "latest", "current", or "up to date" — freshness wording never reroutes documentation questions to web search.
 - Resolve the Context7 library ID first, then query the docs.
 - Do not use Context7 for AWS, Terraform, OpenTofu, or OpenCode documentation.
-- For those exceptions, use the specialized documentation tools instead: `aws-docs`, `terraform-docs`, `opentofu-docs`, and `opencode-docs`.
+- For AWS, Terraform, and OpenTofu documentation, use the specialized tools instead: `aws-docs`, `terraform-docs`, and `opentofu-docs`. For OpenCode configuration, use the checked-in schema and repository validation.
 
 ## MCP integration changes (gateway-first)
 

agents/claude.md

@@ -118,7 +118,7 @@ When asked for a review, adopt a code review mindset:
 - For any operation whose output may exceed ~20 lines, route through `context-mode_ctx_batch_execute` or `context-mode_ctx_execute` so raw output stays in the sandbox.
 - When reading files for analysis rather than editing, prefer `context-mode_ctx_execute_file`. Reading a file you intend to edit is what `read` is for.
 - Use Context7 proactively for current library, framework, SDK, API, CLI, and cloud-service documentation; resolve the library ID first, then query.
-- Prefer dedicated documentation tools over Context7 for AWS, Terraform, OpenTofu, and OpenCode (`aws-docs`, `terraform-docs`, `opentofu-docs`, `opencode-docs`).
+- Prefer dedicated documentation tools over Context7 for AWS, Terraform, and OpenTofu (`aws-docs`, `terraform-docs`, `opentofu-docs`). For OpenCode configuration, use the checked-in schema and repository validation.
 
 ## Skills
 
@@ -140,10 +140,10 @@ This file is one layer in a multi-layer instruction stack. The effective behavio
 - **Tool availability and permissions.** The exact set of available tools depends on MCP server configuration and permission mode. A typical session includes built-in tools (`read`, `edit`, `glob`, `grep`, `bash`, `write`, `task`, `todowrite`), plus GitHub, tmux, and additional MCP servers. Tool calls may require interactive approval, and deferred MCP tools may need a discovery/search step before use.
 - **Context-mode routing.** `AGENTS.md` defines mandatory routing rules that intercept and redirect tool calls to protect the context window. This includes blocking shell HTTP, redirecting large-output operations to sandboxed execution, and enforcing a tool selection hierarchy. This layer fundamentally shapes how tools are used in practice.
 - **Context management.** Automatic conversation compression, context window limits, and output truncation are runtime behaviors outside this file's control.
-- **Memory system.** Persistent cross-session memory (file-based and/or MCP-backed) provides structured storage, recall, and indexing. Its behavior and location depend on runtime configuration, not this file.
+- **Memory system.** Persistent cross-session memory (file-based and/or MCP-backed) provides structured storage, recall, and indexing. Its behavior and location depend on the runtime configuration, not this file.
 - **Skills system.** Loadable skill modules inject domain-specific instructions and workflows on demand. Skills are discovered and loaded at runtime, and the available set is environment-specific.
 - **Subagent system.** The `task` tool launches specialized subagents (typically `explore`, `general`, `bullshit-detector`, `minimax`, plus any repo-defined agents) for parallel research, broad exploration, or delegated work. Availability and capabilities are runtime-dependent.
 - **Scheduling and orchestration.** Recurring tasks, scheduled remote agents, self-paced loops, and deterministic multi-agent workflows are runtime features gated by explicit opt-in and platform support; they are not portable through this file.
 - **Hook-injected guidance.** Session and tool hooks may inject context-window-protection guidance, command-routing tips, and session-specific reminders that override defaults in this file. The exact hook configuration is environment-specific.
-- **Agent hub.** Multi-agent collaboration tools allow registration, messaging, feature planning, and task delegation across concurrent agent sessions. This capability is entirely external to this file.
+- **Agent hub.** Multi-agent collaboration tools allow registration, messaging, feature planning, and service discovery across concurrent agent sessions. This capability is entirely external to this file.
 - **Model capabilities.** Reasoning depth, knowledge cutoff, multimodal understanding, and token limits are properties of the underlying model, not this file.

agents/xnoto.md

@@ -75,7 +75,7 @@ For cross-repo work, summarize the intended order before changing files: source
 - Preserve platform behavior in `.chezmoiignore` and templates: macOS uses `.zprofile`/AeroSpace and excludes Linux shell/i3/systemd paths; Linux uses `.bashrc.d`/i3 and excludes `.zprofile`/AeroSpace.
 - Keep `.chezmoiexternal.toml.tmpl` authoritative for external config repos. If changing how a config repo is fetched, inspect the current type (`git-repo` vs `archive`), target path, branch/archive URL, and refresh behavior.
 - Never copy a rendered external repo wholesale into `dotfiles`. Keep independent repos independent.
-- For OpenCode config changes, prefer the `opencode-docs` MCP over guessing schema or agent file format. Keep `AGENTS.md` context-mode/context7 routing consistent with the runtime config.
+- For OpenCode config changes, use the checked-in `opencode.json` schema and repository validation rather than guessing schema or agent file format. Keep `AGENTS.md` context-mode/context7 routing consistent with the runtime config.
 - OpenCode loads configuration at startup. After changing `opencode.json`, agents, skills, plugins, or MCP configuration, remind the user to restart OpenCode; do not apply or restart it without confirmation.
 
 ---

opencode.json

@@ -146,7 +146,6 @@
     "notion_notion-query-database-view": "allow",
     "notion_notion-query-meeting-notes": "allow",
     "notion_notion-search": "allow",
-    "opencode-docs_*": "allow",
     "opentofu-docs_*": "allow",
     "read": "allow",
     "skill": "allow",
@@ -281,12 +280,6 @@
       "url": "https://mcp.opentofu.org/mcp",
       "enabled": true
     },
-    "opencode-docs": {
-      "type": "remote",
-      "url": "http://127.0.0.1:8773/mcp",
-      "enabled": true,
-      "oauth": false
-    },
     "parallel-search": {
       "type": "remote",
       "url": "https://mcp-makeitwork-parallel-search.makeitwork.cloud/mcp",