@ema.co/mcp-toolkit 2026.1.25 → 2026.1.26-4

package/docs/proposals/explicit-method-restructure.md

@@ -1,328 +0,0 @@
-# MCP Tool Restructure: Explicit Methods
-
-> **Status**: APPROVED - Ready for implementation  
-> **Date**: 2026-01-19  
-> **Branch**: `feat/mcp-tool-restructure`
-
-## Summary
-
-Restructure all MCP tool operations to use explicit `method` parameters instead of inferring operations from the presence of other parameters.
-
-## Motivation
-
-### Current Problem
-
-```typescript
-// What operation is this?
-persona(id="abc", analyze=true)      // analyze? get with flag?
-persona(id="abc", delete=true)       // delete? get with flag?
-persona(id="abc")                    // get? list filtered? 
-persona(data={upload: "path"})       // upload inferred from param presence
-```
-
-The current design:
-1. **Conflates LLM intent with MCP dispatch** - MCP infers what operation to perform
-2. **Ambiguous** - Multiple params present = unclear which operation
-3. **Hard to validate** - Can't validate required params without knowing the operation
-4. **Inconsistent** - Some operations explicit (`create={}`) others implicit (`analyze=true`)
-
-### Principle
-
-```
-LLM = interprets user intent, chooses operation
-MCP = executes structured request
-```
-
-The LLM should explicitly declare what operation it wants. MCP should not guess.
-
-## Design
-
-### Persona Tool
-
-All operations require explicit `method` parameter.
-
-```typescript
-// Query
-persona(method="list")
-persona(method="list", type="voice", status="active")
-persona(method="get", id="abc")
-
-// Mutate
-persona(method="create", name="Bot", type="voice")
-persona(method="create", name="Bot", from="template-id")    // from template
-persona(method="create", name="Bot", from="persona-id")     // clone
-persona(method="update", id="abc", config={...})
-persona(method="update", id="abc", input="add HITL")
-persona(method="delete", id="abc", confirm=true)
-
-// Analyze
-persona(method="analyze", id="abc")
-
-// Sanitize
-persona(method="sanitize", id="abc")                              // preview
-persona(method="sanitize", id="abc", examples=[...], apply=true)  // apply
-
-// History
-persona(method="snapshot", id="abc", message="before change")
-persona(method="history", id="abc")
-persona(method="restore", id="abc", version="v3")
-
-// Compare
-persona(method="compare", id="abc", to="def")
-```
-
-### Data Sub-Resource
-
-Data operations also require explicit `method` inside the `data` object.
-
-```typescript
-// Discovery (MCP provides info to LLM)
-persona(id="abc", data={method:"list"})
-persona(id="abc", data={method:"schema"})   // Get input schema for LLM to generate content
-
-// Create (LLM provides content)
-persona(id="abc", data={method:"upload", path:"/file.pdf"})
-persona(id="abc", data={method:"upload", content:[{name:"doc.md", document:"..."}]})
-
-// Copy (deterministic operation)
-persona(id="abc", data={method:"copy", from:"source-id"})
-persona(id="abc", data={method:"copy", from:"source-id", sanitize:true})
-
-// Manage
-persona(id="abc", data={method:"delete", id:"item-123"})
-persona(id="abc", data={method:"embed", enabled:true})
-persona(id="abc", data={method:"search", query:"pricing"})
-```
-
-### Catalog Tool
-
-Already has `type` as required discriminator. Add `method` for consistency:
-
-```typescript
-catalog(method="list", type="actions")
-catalog(method="get", type="actions", id="knowledge_search")
-catalog(method="search", type="actions", query="email")
-catalog(method="recommend", type="actions", for="send email with approval")
-```
-
-### Sync Tool
-
-```typescript
-sync(method="preview", persona="abc", from="staging", to="prod")
-sync(method="execute", persona="abc", from="staging", to="prod")
-sync(method="status")
-sync(method="status", persona="abc")
-```
-
-## Schema
-
-### Persona Tool Schema
-
-```typescript
-{
-  name: "persona",
-  inputSchema: {
-    type: "object",
-    properties: {
-      // Required method
-      method: {
-        type: "string",
-        enum: ["list", "get", "create", "update", "delete", 
-               "analyze", "sanitize", "snapshot", "history", "restore", "compare"],
-        description: "Operation to perform (required)"
-      },
-      
-      // Identity
-      id: { type: "string", description: "Persona ID or name" },
-      
-      // Filters (for list)
-      type: { type: "string", enum: ["voice", "chat", "dashboard"] },
-      status: { type: "string", enum: ["active", "inactive", "disabled"] },
-      query: { type: "string", description: "Search by name" },
-      
-      // Create params
-      name: { type: "string", description: "Name (for method=create)" },
-      from: { type: "string", description: "Template/persona ID to clone from" },
-      
-      // Update params
-      config: { type: "object", description: "Proto config" },
-      input: { type: "string", description: "Natural language instruction" },
-      workflow_spec: { type: "object", description: "WorkflowSpec (agent-built)" },
-      
-      // Delete params
-      confirm: { type: "boolean", description: "Confirm deletion" },
-      
-      // Sanitize params
-      examples: { type: "array", items: { type: "string" } },
-      apply: { type: "boolean", description: "Apply sanitization" },
-      
-      // Snapshot params
-      message: { type: "string", description: "Snapshot message" },
-      
-      // Restore params
-      version: { type: "string", description: "Version to restore" },
-      
-      // Compare params
-      to: { type: "string", description: "Compare target ID" },
-      
-      // Options
-      include_workflow: { type: "boolean" },
-      preview: { type: "boolean" },
-      env: { type: "string" },
-      
-      // Data sub-resource
-      data: {
-        type: "object",
-        properties: {
-          method: {
-            type: "string",
-            enum: ["list", "schema", "upload", "copy", "delete", "embed", "search"],
-            description: "Data operation (required)"
-          },
-          from: { type: "string", description: "Source persona (for copy)" },
-          sanitize: { type: "boolean", description: "Mask PII (for copy)" },
-          path: { type: "string", description: "File path (for upload)" },
-          content: { type: "array", items: { type: "object" }, description: "Content array (for upload)" },
-          id: { type: "string", description: "Item ID (for delete)" },
-          enabled: { type: "boolean", description: "Enable/disable (for embed)" },
-          query: { type: "string", description: "Search query" },
-        },
-        required: ["method"],
-      },
-    },
-    required: ["method"],
-  }
-}
-```
-
-## Migration
-
-### Breaking Changes
-
-| Old | New |
-|-----|-----|
-| `persona()` | `persona(method="list")` |
-| `persona(id="abc")` | `persona(method="get", id="abc")` |
-| `persona(analyze=true, id="abc")` | `persona(method="analyze", id="abc")` |
-| `persona(delete=true, id="abc")` | `persona(method="delete", id="abc", confirm=true)` |
-| `persona(sanitize=true, id="abc")` | `persona(method="sanitize", id="abc")` |
-| `persona(data={upload:"path"})` | `persona(id="abc", data={method:"upload", path:"..."})` |
-| `persona(data={dashboard_clone:"src"})` | `persona(id="abc", data={method:"copy", from:"src"})` |
-| `persona(data={dashboard_generate:5})` | REMOVED - LLM generates, uses `data={method:"upload", content:[...]}` |
-
-### Removed Parameters
-
-| Removed | Replacement |
-|---------|-------------|
-| `analyze: boolean` | `method="analyze"` |
-| `delete: boolean` | `method="delete"` |
-| `sanitize: boolean` | `method="sanitize"` |
-| `sanitize_examples` | `examples` (with method="sanitize") |
-| `snapshot: string` | `message` (with method="snapshot") |
-| `history: boolean` | `method="history"` |
-| `restore: string` | `version` (with method="restore") |
-| `compare: string` | `to` (with method="compare") |
-| `dashboard_clone` | `data={method:"copy", from:"..."}` |
-| `dashboard_generate` | `data={method:"upload", content:[...]}` |
-| `generate` (for data) | REMOVED - LLM provides content via upload |
-| `template` (for data) | REMOVED - LLM decides content type |
-
-## Handler Dispatch
-
-```typescript
-export async function handlePersona(args: Args, client: EmaClient): Promise<HandlerResult> {
-  // Data sub-resource
-  if (args.data) {
-    return handlePersonaData(args.id, args.data, client);
-  }
-  
-  // Main dispatch
-  switch (args.method) {
-    case "list":     return handleList(args, client);
-    case "get":      return handleGet(args.id, args, client);
-    case "create":   return handleCreate(args, client);
-    case "update":   return handleUpdate(args, client);
-    case "delete":   return handleDelete(args.id, args.confirm, client);
-    case "analyze":  return handleAnalyze(args.id, args, client);
-    case "sanitize": return handleSanitize(args, client);
-    case "snapshot": return handleSnapshot(args.id, args.message, client);
-    case "history":  return handleHistory(args.id, client);
-    case "restore":  return handleRestore(args.id, args.version, client);
-    case "compare":  return handleCompare(args.id, args.to, client);
-    default:
-      return { error: `Unknown method: ${args.method}`, hint: "Valid: list, get, create, update, delete, analyze, sanitize, snapshot, history, restore, compare" };
-  }
-}
-
-function handlePersonaData(personaId: string, data: DataArgs, client: EmaClient): Promise<HandlerResult> {
-  switch (data.method) {
-    case "list":   return handleDataList(personaId, client);
-    case "schema": return handleDataSchema(personaId, client);
-    case "upload": return handleDataUpload(personaId, data, client);
-    case "copy":   return handleDataCopy(personaId, data, client);
-    case "delete": return handleDataDelete(personaId, data.id, client);
-    case "embed":  return handleDataEmbed(personaId, data.enabled, client);
-    case "search": return handleDataSearch(personaId, data.query, client);
-    default:
-      return { error: `Unknown data method: ${data.method}`, hint: "Valid: list, schema, upload, copy, delete, embed, search" };
-  }
-}
-```
-
-## LLM/MCP Responsibility Split
-
-| Responsibility | LLM | MCP |
-|----------------|-----|-----|
-| Interpret user intent | ✅ | |
-| Choose operation (method) | ✅ | |
-| Generate content | ✅ | |
-| Decide content type | ✅ | |
-| Provide schema | | ✅ |
-| Execute API calls | | ✅ |
-| Apply sanitization | | ✅ |
-| Return structured results | | ✅ |
-
-## Example Flow
-
-```
-User: "Add 5 test contracts to my Contract Analyzer"
-         │
-         ▼
-┌─────────────────────────────────────────────────────┐
-│ LLM                                                 │
-│                                                     │
-│ 1. Understands: user wants test data for dashboard  │
-│ 2. Calls: persona(method="get", id="Contract...")   │
-│ 3. Sees: dashboard type, gets ID                    │
-│ 4. Calls: persona(id="abc", data={method:"schema"}) │
-│ 5. Sees: columns = [document, title, date, value]   │
-│ 6. Generates: 5 realistic contracts internally      │
-│ 7. Calls: persona(id="abc", data={                  │
-│      method:"upload",                               │
-│      content:[                                      │
-│        {document:{...}, title:"...", ...},          │
-│        ...                                          │
-│      ]                                              │
-│    })                                               │
-└─────────────────────────────────────────────────────┘
-         │
-         ▼
-┌─────────────────────────────────────────────────────┐
-│ MCP                                                 │
-│                                                     │
-│ 1. Validates: method="upload", content provided     │
-│ 2. For each item in content:                        │
-│    - Formats for API                                │
-│    - Calls uploadAndRunDashboardRow()               │
-│ 3. Returns: {uploaded: 5, rows: [...]}              │
-└─────────────────────────────────────────────────────┘
-```
-
-## Approval
-
-- [x] Design reviewed
-- [x] Breaking changes documented
-- [x] Migration path clear
-- [ ] Implementation planned
-- [ ] Tests updated
-- [ ] Documentation updated

package/docs/proposals/mcp-tool-restructure-2026-01.md

@@ -1,366 +0,0 @@
-# Proposal: MCP Tool Structure Refactoring
-
-**Date**: 2026-01-17  
-**Status**: Draft  
-**Author**: AI Assistant (with user guidance)
-
-## Executive Summary
-
-Refactor the Ema MCP server from 7 mode-based tools with 30+ parameters each to ~32 focused tools following industry-standard `{resource}_{action}` naming. This aligns with the MCP specification, popular MCP servers (Playwright, Firecrawl, Supabase), and improves LLM discoverability.
-
----
-
-## Problem Statement
-
-### Current State
-
-The Ema MCP server consolidates ~45 legacy tools into 7 "super tools" using a `mode` parameter:
-
-```typescript
-// Current: Mode-based approach
-persona(mode="list")
-persona(mode="get", id="abc")
-persona(mode="create", name="Bot", type="voice", input="...", preview=false)
-persona(mode="analyze", id="abc", fix=true, include=["workflow", "config"])
-```
-
-### Issues
-
-| Issue | Impact |
-|-------|--------|
-| **30+ params per tool** | LLMs see all params even when irrelevant to the mode |
-| **Mode confusion** | "Did I need `mode='get'` or just pass `id`?" |
-| **Inconsistent patterns** | Some tools use `mode`, others use flags (`all=true`) |
-| **Overlapping functionality** | `reference(type="actions")` duplicates `action(all=true)` |
-| **3 deprecated tools visible** | `workflow`, `knowledge`, `demo` add noise |
-
-### Evidence
-
-- **Playwright MCP**: Uses `browser_click`, `browser_navigate`, `browser_type` (15+ focused tools)
-- **Firecrawl MCP**: Uses `firecrawl_scrape`, `firecrawl_crawl`, `firecrawl_search` (6 tools)
-- **Filesystem MCP**: Uses `read_file`, `write_file`, `list_directory` (10+ tools)
-- **MCP Spec**: Explicitly allows `admin.tools.list` style naming; examples use `get_weather`
-
----
-
-## Proposed Solution
-
-### Tool Naming Convention
-
-**Pattern**: `{resource}_{action}` (snake_case)
-
-```typescript
-// Proposed: Action-based approach
-persona_list()
-persona_get(id="abc")
-persona_create(name="Bot", type="voice", input="...")
-persona_analyze(id="abc", fix=true)
-```
-
-### Why This Pattern
-
-1. **Alphabetical grouping**: `persona_*` tools cluster together in tool lists
-2. **Industry standard**: Matches Playwright, Firecrawl, Filesystem servers
-3. **MCP spec compliant**: Allowed chars include underscore, examples show similar patterns
-4. **Focused parameters**: Each tool only shows relevant options
-5. **No mode confusion**: Operation is in the name
-
----
-
-## Proposed Tool Set
-
-### Core Tools (~32 total)
-
-#### Persona Operations (8 tools)
-
-| Tool | Description | Key Params |
-|------|-------------|------------|
-| `persona_list` | List AI Employees | `query?`, `status?`, `type?`, `limit?` |
-| `persona_get` | Get specific persona | `id`, `include_workflow?` |
-| `persona_create` | Create new persona | `name`, `type\|from`, `input?`, `preview?` |
-| `persona_update` | Update config | `id`, `proto_config?`, `workflow?` |
-| `persona_clone` | Clone persona | `from`, `name`, `include_data?`, `sanitize?` |
-| `persona_analyze` | Analyze for issues | `id`, `fix?`, `include?` |
-| `persona_sanitize` | Remove PII | `id`, `sanitize_examples?` |
-| `persona_templates` | List templates | (none) |
-
-#### Version Operations (4 tools)
-
-| Tool | Description | Key Params |
-|------|-------------|------------|
-| `version_create` | Create snapshot | `persona_id`, `message?` |
-| `version_list` | List versions | `persona_id` |
-| `version_restore` | Restore version | `persona_id`, `version` |
-| `version_policy` | Set version policy | `persona_id`, `auto_on_deploy?` |
-
-#### Data Operations (7 tools)
-
-| Tool | Description | Key Params |
-|------|-------------|------------|
-| `data_list` | List files | `persona_id`, `limit?` |
-| `data_get` | Get specific file | `persona_id`, `file_id` |
-| `data_upload` | Upload file | `persona_id`, `file`, `tags?` |
-| `data_generate` | Generate content | `persona_id`, `from\|input`, `count?` |
-| `data_delete` | Delete file | `persona_id`, `file_id` |
-| `data_sanitize` | Sanitize data | `persona_id`, `include_workflow?` |
-| `data_embedding` | Embedding status/toggle | `persona_id`, `enabled?` |
-
-#### Action Operations (4 tools)
-
-| Tool | Description | Key Params |
-|------|-------------|------------|
-| `action_list` | List workflow actions | `category?`, `limit?` |
-| `action_get` | Get action details | `id`, `include_docs?` |
-| `action_suggest` | Suggest for use case | `input` |
-| `action_categories` | List categories | (none) |
-
-#### Template Operations (4 tools)
-
-| Tool | Description | Key Params |
-|------|-------------|------------|
-| `template_patterns` | List workflow patterns | `type?` |
-| `template_pattern` | Get specific pattern | `name` |
-| `template_widgets` | Get widget reference | `type` |
-| `template_questions` | Get qualifying questions | `category?` |
-
-#### Reference Operations (5 tools)
-
-| Tool | Description | Key Params |
-|------|-------------|------------|
-| `reference_concept` | Get concept definition | `term` |
-| `reference_concepts` | List all concepts | (none) |
-| `reference_guidance` | Get topic guidance | `topic` |
-| `reference_validate` | Validate types/prompt | `check_types?`, `validate_prompt?` |
-| `reference_debug` | Get debug info | `type` (mistakes/checklist/execution) |
-
-#### Sync Operations (3 tools)
-
-| Tool | Description | Key Params |
-|------|-------------|------------|
-| `sync_run` | Sync to target env | `id?`, `target`, `scope?`, `dry_run?` |
-| `sync_status` | Check sync status | `id?`, `list_synced?` |
-| `sync_config` | Show sync config | (none) |
-
-#### Environment (1 tool)
-
-| Tool | Description | Key Params |
-|------|-------------|------------|
-| `env` | List environments | (none) |
-
----
-
-## Prompts (User-Controlled Workflows)
-
-Per MCP spec, prompts are for **user-initiated workflows**, not model-controlled operations.
-
-### Proposed Prompts
-
-| Prompt | Description | Arguments |
-|--------|-------------|-----------|
-| `create_voice_ai` | Guided Voice AI creation | `use_case`, `name?` |
-| `create_chat_ai` | Guided Chat AI creation | `use_case`, `name?` |
-| `clone_for_demo` | Clone + sanitize for demo | `source_id` |
-| `troubleshoot` | Diagnose workflow issues | `persona_id` |
-| `sync_to_production` | Safe production sync | `persona_id` |
-| `migrate_persona` | Cross-env migration guide | `persona_id`, `target_env` |
-
----
-
-## Migration Strategy
-
-### Phase 1: Add New Tools (Non-Breaking)
-
-1. Add all `{resource}_{action}` tools alongside existing tools
-2. Both patterns work simultaneously
-3. Document new pattern as preferred
-
-### Phase 2: Deprecate Mode-Based Tools
-
-1. Add deprecation warnings to mode-based tools
-2. Route mode-based calls to new tools internally
-3. Update documentation and examples
-
-### Phase 3: Remove Deprecated Tools
-
-1. Remove `workflow`, `knowledge`, `demo` (already deprecated)
-2. Remove mode-based `persona`, `data`, `action`, etc.
-3. Final tool count: ~32
-
-### Backwards Compatibility
-
-```typescript
-// Internal routing for backwards compatibility
-if (toolName === "persona" && args.mode) {
-  // Route to new tool
-  const newTool = `persona_${args.mode}`;
-  console.warn(`Deprecated: Use ${newTool} instead of persona(mode="${args.mode}")`);
-  return handleTool(newTool, args);
-}
-```
-
----
-
-## Implementation Details
-
-### Tool Definition Example
-
-```typescript
-// tools.ts
-
-export const TOOLS: Tool[] = [
-  {
-    name: "persona_list",
-    description: "List AI Employees with optional filters",
-    inputSchema: {
-      type: "object",
-      properties: {
-        query: { type: "string", description: "Search by name" },
-        status: { type: "string", enum: ["active", "inactive", "draft"] },
-        type: { type: "string", enum: ["voice", "chat", "dashboard"] },
-        limit: { type: "number", description: "Max results (default: 50)" },
-        env: { type: "string", description: "Target environment" },
-      },
-    },
-  },
-  {
-    name: "persona_get",
-    description: "Get a specific AI Employee by ID",
-    inputSchema: {
-      type: "object",
-      properties: {
-        id: { type: "string", description: "Persona ID or name" },
-        include_workflow: { type: "boolean", description: "Include full workflow" },
-        env: { type: "string" },
-      },
-      required: ["id"],
-    },
-  },
-  // ... etc
-];
-```
-
-### Handler Routing
-
-```typescript
-// server.ts
-
-server.setRequestHandler(CallToolRequestSchema, async (request) => {
-  const { name, arguments: args } = request.params;
-  
-  // Parse tool name: "persona_list" → resource="persona", action="list"
-  const [resource, action] = name.split("_");
-  
-  switch (resource) {
-    case "persona":
-      return handlePersona(action, args);
-    case "data":
-      return handleData(action, args);
-    case "action":
-      return handleAction(action, args);
-    case "template":
-      return handleTemplate(action, args);
-    case "reference":
-      return handleReference(action, args);
-    case "sync":
-      return handleSync(action, args);
-    case "version":
-      return handleVersion(action, args);
-    case "env":
-      return handleEnv(args);
-    default:
-      throw new Error(`Unknown tool: ${name}`);
-  }
-});
-```
-
----
-
-## Comparison Matrix
-
-| Aspect | Current (mode-based) | Proposed (action-based) |
-|--------|---------------------|------------------------|
-| Tool count | 7 (+ 3 deprecated) | ~32 |
-| Params per tool | 30+ | 3-8 |
-| Discoverability | Poor (read long descriptions) | Excellent (scan tool names) |
-| LLM clarity | Confusing modes | Clear operations |
-| Industry alignment | Non-standard | Standard (Playwright, Firecrawl) |
-| MCP spec alignment | Partial | Full |
-| Alphabetical grouping | No | Yes (`persona_*` clusters) |
-
----
-
-## Risks & Mitigations
-
-| Risk | Mitigation |
-|------|------------|
-| Breaking existing integrations | Phased rollout with backwards compat routing |
-| Tool count explosion | Capped at ~32, well-organized by prefix |
-| Documentation churn | Generate docs from tool definitions |
-| LLM token cost | More tools but simpler schemas (net neutral) |
-
----
-
-## Success Metrics
-
-1. **LLM accuracy**: Fewer incorrect tool calls
-2. **Tool discovery**: Users find tools faster in lists
-3. **Param errors**: Fewer "wrong param for this mode" issues
-4. **Adoption**: New pattern used in 90%+ of calls within 30 days
-
----
-
-## Files to Modify
-
-| File | Changes |
-|------|---------|
-| `src/mcp/tools-consolidated.ts` | Replace with new tool definitions |
-| `src/mcp/server.ts` | Update handler routing |
-| `src/mcp/prompts.ts` | Add workflow prompts |
-| `docs/mcp-tools-guide.md` | Document new pattern |
-| `test/mcp-tools.test.ts` | Update tests |
-
----
-
-## Timeline
-
-| Phase | Scope | 
-|-------|-------|
-| Phase 1 | Add new tools alongside existing |
-| Phase 2 | Deprecate mode-based tools |
-| Phase 3 | Remove deprecated tools |
-
----
-
-## Appendix: Tool Name Mapping
-
-| Current | New |
-|---------|-----|
-| `persona(mode="list")` | `persona_list` |
-| `persona(mode="get", id="...")` | `persona_get` |
-| `persona(mode="create", ...)` | `persona_create` |
-| `persona(mode="update", ...)` | `persona_update` |
-| `persona(mode="clone", ...)` | `persona_clone` |
-| `persona(mode="analyze", ...)` | `persona_analyze` |
-| `persona(mode="sanitize", ...)` | `persona_sanitize` |
-| `persona(mode="templates")` | `persona_templates` |
-| `data(mode="list", ...)` | `data_list` |
-| `data(mode="upload", ...)` | `data_upload` |
-| `data(mode="generate", ...)` | `data_generate` |
-| `data(mode="delete", ...)` | `data_delete` |
-| `action(all=true)` | `action_list` |
-| `action(id="...")` | `action_get` |
-| `action(suggest="...")` | `action_suggest` |
-| `sync(mode="run", ...)` | `sync_run` |
-| `sync(mode="status", ...)` | `sync_status` |
-| `reference(concept="...")` | `reference_concept` |
-| `reference(guidance="...")` | `reference_guidance` |
-
----
-
-## References
-
-- [MCP Specification - Tools](https://modelcontextprotocol.io/specification/2025-11-25/server/tools)
-- [MCP Specification - Prompts](https://modelcontextprotocol.io/specification/2025-11-25/server/prompts)
-- [Playwright MCP Server](https://github.com/microsoft/playwright-mcp)
-- [Firecrawl MCP Server](https://github.com/firecrawl/firecrawl-mcp-server)
-- [Awesome MCP Servers](https://mcpservers.org/)