{"_id":"@ai-mantras/mcp-server","name":"@ai-mantras/mcp-server","dist-tags":{"latest":"1.3.0"},"versions":{"1.3.0":{"name":"@ai-mantras/mcp-server","version":"1.3.0","description":"MCP server for the AI Mantras prompt framework - personas, patterns, and orchestration for Claude","type":"module","main":"dist/index.js","bin":{"ai-mantras-mcp":"dist/index.js"},"scripts":{"copy-content":"node scripts/copy-content.js","prebuild":"npm run copy-content","build":"tsc","start":"node dist/index.js","dev":"tsx src/index.ts","test":"node test-content.js && node test-tools.js && node test-multi-agent.js"},"keywords":["mcp","claude","ai","personas","patterns","prompts","orchestration"],"author":{"name":"James Bennett"},"license":"MPL-2.0","dependencies":{"@anthropic-ai/sdk":"^0.39.0","@modelcontextprotocol/sdk":"^1.0.0","openai":"^4.70.0","uuid":"^9.0.0","yaml":"^2.3.4"},"devDependencies":{"@types/node":"^20.10.0","@types/uuid":"^9.0.0","tsx":"^4.7.0","typescript":"^5.3.0"},"repository":{"type":"git","url":"git+https://github.com/kd5ziy/AIMantras.git","directory":"ai-mantras-mcp"},"homepage":"https://aimantras.org","bugs":{"url":"https://github.com/kd5ziy/AIMantras/issues"},"publishConfig":{"access":"public"},"engines":{"node":">=18.0.0"},"_id":"@ai-mantras/mcp-server@1.3.0","gitHead":"55a430acf5678d335f56a56fd0713f32cff37aa0","types":"./dist/index.d.ts","_nodeVersion":"22.21.1","_npmVersion":"10.9.4","dist":{"integrity":"sha512-cha3RafgmL1QhMs+X0Tg9gb/jDTNo15w1ABM/0ez1mmGfaRb+QgPwtG/qEbUtWG4Wvb22X8QkDRobUIDmx7ejg==","shasum":"a4415f44828a34de8845fccf0f12bb6bd0af8249","tarball":"https://registry.npmjs.org/@ai-mantras/mcp-server/-/mcp-server-1.3.0.tgz","fileCount":127,"unpackedSize":471159,"signatures":[{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEQCIHNgYcbRA9spt5eYdRKtpA3Kr9W5GC6k8p9eMUwGNX99AiBUSE3PLK30jkLMdVKQjcL5k8kKTM1Ts1ddexLXnK3PNQ=="}]},"_npmUser":{"name":"kd5ziy","email":"kd5ziy@gmail.com"},"directories":{},"maintainers":[{"name":"kd5ziy","email":"kd5ziy@gmail.com"}],"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/mcp-server_1.3.0_1771221364511_0.5541247102652254"},"_hasShrinkwrap":false}},"time":{"created":"2026-02-16T05:56:04.386Z","1.3.0":"2026-02-16T05:56:04.663Z","modified":"2026-02-16T05:56:04.889Z"},"maintainers":[{"name":"kd5ziy","email":"kd5ziy@gmail.com"}],"description":"MCP server for the AI Mantras prompt framework - personas, patterns, and orchestration for Claude","homepage":"https://aimantras.org","keywords":["mcp","claude","ai","personas","patterns","prompts","orchestration"],"repository":{"type":"git","url":"git+https://github.com/kd5ziy/AIMantras.git","directory":"ai-mantras-mcp"},"author":{"name":"James Bennett"},"bugs":{"url":"https://github.com/kd5ziy/AIMantras/issues"},"license":"MPL-2.0","readme":"# AI Mantras MCP Server\r\n\r\nAn MCP (Model Context Protocol) server that exposes the AI Mantras prompt framework as structured tools, resources, and prompts for Claude.\r\n\r\n## What is AI Mantras?\r\n\r\nAI Mantras is a prompt engineering framework that treats AI collaboration as organizational design. It provides:\r\n\r\n- **11 Personas** across orchestration, domain, and evaluation categories\r\n- **9 Thinking Patterns** for structured reasoning\r\n- **24 Skills** - actionable capabilities personas can invoke\r\n- **3 Complexity Tiers** (Simple, Moderate, Complex) with appropriate workflows\r\n- **Separation of Powers** to prevent self-evaluation and ensure quality\r\n\r\n## Installation\r\n\r\n### For Users (Zero Config)\r\n\r\n```bash\r\n# Install globally\r\nnpm install -g @ai-mantras/mcp-server\r\n\r\n# Or use with npx (no install needed)\r\n```\r\n\r\nAdd to your Claude Code `.mcp.json`:\r\n\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"ai-mantras\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"@ai-mantras/mcp-server\"]\r\n    }\r\n  }\r\n}\r\n```\r\n\r\n### For Development\r\n\r\n```bash\r\ncd ai-mantras-mcp\r\nnpm install\r\nnpm run build\r\n```\r\n\r\n## Tools\r\n\r\n### Core Tools\r\n\r\n| Tool | Description |\r\n|------|-------------|\r\n| `bootstrap_session` | Initialize a session with appropriate resources for a request |\r\n| `assess_complexity` | Triage a request into Simple/Moderate/Complex tier |\r\n| `get_persona` | Load a persona by name or domain match |\r\n| `get_pattern` | Load a thinking pattern by name |\r\n| `get_skill` | Load a skill by name, task match, or category |\r\n| `get_workflow` | Get workflow steps for a complexity tier |\r\n| `create_handoff` | Generate handoff template between personas |\r\n| `list_available` | List all available personas, patterns, and skills |\r\n\r\n### Multi-Agent Tools\r\n\r\nThese tools enable true multi-agent execution with context isolation between personas.\r\n\r\n> **Note:** These tools only appear when `MANTRAS_MULTI_AGENT_ENABLED=true`\r\n\r\n| Tool | Description |\r\n|------|-------------|\r\n| `spawn_agent` | Spawn an isolated AI agent with a specific persona |\r\n| `get_agent_result` | Retrieve status and result from a spawned agent |\r\n| `list_agents` | List all spawned agents with status and usage stats |\r\n\r\n#### spawn_agent\r\n\r\nSpawns an isolated agent that only sees its persona definition, the task, and any provided inputs. Supports both synchronous (wait for result) and asynchronous (return immediately) modes.\r\n\r\n```javascript\r\nspawn_agent({\r\n  persona: \"Clara-Financial-Analyst\",  // Required: persona name\r\n  task: \"Analyze this investment opportunity\", // Required: what to do\r\n  inputs: { document: \"...\" },          // Optional: context data\r\n  success_criteria: \"Provide ROI estimate\", // Optional: success definition\r\n  patterns: [\"chain-of-thought\"],       // Optional: thinking patterns\r\n  provider: \"anthropic\",                // Optional: \"anthropic\" or \"openai\"\r\n  model: \"claude-sonnet-4-20250514\",   // Optional: specific model\r\n  async: false,                         // Optional: async mode\r\n  max_tokens: 4096,                     // Optional: response limit\r\n  timeout_ms: 120000                    // Optional: timeout\r\n})\r\n```\r\n\r\n#### get_agent_result\r\n\r\nPoll for completion or retrieve results from a spawned agent.\r\n\r\n```javascript\r\nget_agent_result({\r\n  agent_id: \"agent_abc12345\"  // Required: ID from spawn_agent\r\n})\r\n```\r\n\r\n#### list_agents\r\n\r\nList all spawned agents with optional status filtering.\r\n\r\n```javascript\r\nlist_agents({\r\n  status: \"all\",  // Optional: pending|running|completed|failed|timeout|all\r\n  limit: 20       // Optional: max results\r\n})\r\n```\r\n\r\n## Resources\r\n\r\n| URI | Content |\r\n|-----|---------|\r\n| `mantras://principles` | Guiding principles |\r\n| `mantras://bootstrapper` | Agent bootstrapper |\r\n| `mantras://manifest` | Framework manifest (YAML) |\r\n| `mantras://toolset` | Skills master index |\r\n| `mantras://persona/{category}/{name}` | Individual personas |\r\n| `mantras://pattern/{layer}/{name}` | Individual patterns |\r\n| `mantras://skill/{category}/{name}` | Individual skills |\r\n\r\n## Prompts\r\n\r\n| Prompt | Description |\r\n|--------|-------------|\r\n| `mantras` | Auto-triage and process request |\r\n| `mantras-simple` | Direct answer with single persona |\r\n| `mantras-moderate` | Structured reasoning + self-review |\r\n| `mantras-complex` | Full orchestration workflow |\r\n\r\n## Usage Examples\r\n\r\n### Using the main prompt\r\n```\r\n/mantras \"What's the best filesystem for a NAS?\"\r\n```\r\n\r\n### Using tools directly\r\n```\r\nUse the assess_complexity tool to analyze: \"Help me design a backup strategy\"\r\n```\r\n\r\n### Reading resources\r\n```\r\nRead mantras://persona/domain/Kestra-Systems-Architect\r\n```\r\n\r\n## Environment Variables\r\n\r\n### Operating Modes\r\n\r\nAI Mantras supports two operating modes:\r\n\r\n| Mode | Description | Use Case |\r\n|------|-------------|----------|\r\n| **Single-Agent** (default) | All personas run on one AI model | Simple setup, no API keys needed, lower cost |\r\n| **Multi-Agent** | Each persona spawns as a separate isolated agent | True context isolation, parallel execution, model specialization |\r\n\r\nSet the mode via environment variable:\r\n\r\n```bash\r\nMANTRAS_MULTI_AGENT_ENABLED=false  # Single-agent mode (default)\r\nMANTRAS_MULTI_AGENT_ENABLED=true   # Multi-agent mode\r\n```\r\n\r\n### Multi-Agent Configuration\r\n\r\nWhen `MANTRAS_MULTI_AGENT_ENABLED=true`, configure these additional settings:\r\n\r\n| Variable | Description | Required |\r\n|----------|-------------|----------|\r\n| `MANTRAS_MULTI_AGENT_ENABLED` | Enable multi-agent tools | No (default: false) |\r\n| `ANTHROPIC_API_KEY` | Anthropic API key for Claude models | For Anthropic provider |\r\n| `OPENAI_API_KEY` | OpenAI API key for GPT models | For OpenAI provider |\r\n| `MANTRAS_DEFAULT_PROVIDER` | Default provider: `anthropic` or `openai` | No (default: anthropic) |\r\n| `MANTRAS_DEFAULT_MODEL` | Default model name | No (auto-selected) |\r\n| `MANTRAS_MAX_CONCURRENT_AGENTS` | Max parallel async agents | No (default: 5) |\r\n| `MANTRAS_AGENT_TIMEOUT_MS` | Default timeout in ms | No (default: 120000) |\r\n\r\n### Example Configurations\r\n\r\n**Single-Agent Mode** (default, no extra config needed):\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"ai-mantras\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"@ai-mantras/mcp-server\"]\r\n    }\r\n  }\r\n}\r\n```\r\n\r\n**Multi-Agent Mode** (enables spawn_agent, get_agent_result, list_agents):\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"ai-mantras\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"@ai-mantras/mcp-server\"],\r\n      \"env\": {\r\n        \"MANTRAS_MULTI_AGENT_ENABLED\": \"true\",\r\n        \"ANTHROPIC_API_KEY\": \"sk-ant-...\"\r\n      }\r\n    }\r\n  }\r\n}\r\n```\r\n\r\n## Custom Content\r\n\r\nBy default, the server uses bundled content. Power users can override with custom content:\r\n\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"ai-mantras\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"@ai-mantras/mcp-server\"],\r\n      \"env\": {\r\n        \"MANTRAS_CONTENT_PATH\": \"/path/to/your/content\"\r\n      }\r\n    }\r\n  }\r\n}\r\n```\r\n\r\n## Available Personas\r\n\r\n### Orchestration\r\n- **Bernstein-Orchestrator** - Coordinates multi-persona workflows\r\n- **Hopper-Project-Planner** - Frames problems and decomposes work\r\n- **Lovell-Crisis-Planner** - Handles high-stakes, time-sensitive planning\r\n\r\n### Domain\r\n- **Clara-Financial-Analyst** - Financial analysis and portfolio strategy\r\n- **Kestra-Systems-Architect** - Infrastructure and systems design\r\n- **Goeth-Philosophical-Synthesizer** - Ethics and meaning frameworks\r\n- **Franklin-Deep-Reasoner** - Multi-domain integration\r\n- **Watson-Medical-Advisor** - Evidence-based medical guidance\r\n\r\n### Evaluation\r\n- **Ada-QA-Reviewer** - Quality assurance and review\r\n- **Drucker-Goal-Satisfaction-Evaluator** - Objective measurement\r\n- **Rickover-Safety-Evaluator** - Risk and safety assessment\r\n\r\n## Available Patterns\r\n\r\n### Layer 1: Foundational\r\n- `planning-phase` - Problem definition before execution\r\n- `orchestration` - Multi-persona coordination\r\n- `recursive-self-eval` - Critique and refinement loops\r\n- `meta-rules` - Behavioral guardrails\r\n\r\n### Layer 2: Thinking Primitives\r\n- `chain-of-thought` - Labeled reasoning stages\r\n- `rule-based-reasoning` - Explicit rules with traceability\r\n- `guardrail-creative` - Creativity within constraints\r\n\r\n### Layer 3: Evaluation\r\n- `criterion-based-evaluation` - Systematic assessment\r\n- `threat-modeling` - Risk identification\r\n\r\n## Available Skills\r\n\r\n### Research\r\n- `web-search` - Search and synthesize web information\r\n- `codebase-exploration` - Navigate and understand code\r\n\r\n### Analysis\r\n- `code-review` - Review code for quality and issues\r\n- `risk-assessment` - Evaluate risks and mitigations\r\n\r\n### Creation\r\n- `document-generation` - Create structured documents\r\n- `code-generation` - Generate code from specifications\r\n\r\n### Evaluation\r\n- `quality-check` - Verify output quality\r\n- `security-audit` - Assess security posture\r\n\r\n### Orchestration\r\n- `task-decomposition` - Break down complex tasks\r\n- `handoff` - Transfer work between personas\r\n\r\n### Utility\r\n- `help` - Framework guidance and discovery\r\n\r\n## License\r\n\r\nMPL-2.0\r\n","readmeFilename":"README.md","_rev":"1-c583f29eee1147eadfaeaad44e93e688"}