# Strategic Agent Reasoning MCP

> Strategic Agent Reasoning MCP is an open-source MCP server for ai models and llm tooling. Belief-desire-intention reasoning engine for agents: goals, utility scoring and replanning. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted
- **Website:** https://agentreasoningmcp.com
- **Repository:** https://github.com/putervision/agent-reasoning-mcp
- **MCP Registry name:** io.github.putervision/agent-reasoning-mcp
- **Categories:** AI models and LLM tooling, developer tools
- **Listing updated:** 2026-11-16

## About

This server gives an agent an explicit decision loop rather than leaving planning implicit in a prompt. Beliefs carry confidence that decays exponentially and is swept by time-to-live, goals decompose into directed acyclic graphs with progress rolling up from the leaves, and candidate actions are ranked by multi-attribute expected utility over weights for aggression, caution, greed, exploration, efficiency and cooperation. Each proposal gets a risk score covering failure probability, reversibility and worst-case blast radius.

The 15 tools split into ten deliberation tools and five fast decision primitives. The fast layer runs locally with in-memory caches and no external calls: classify, ask_noul, ask_choice, ask_score and a gate_intention check that issues signed HMAC dispatch tokens. Its README reports cache-hit latencies in microseconds on Node 22, and it credits the typed pattern to TypeSafe's Jev.

Selected actions become intentions queued for runtime engines, with explainable decision traces and a SHA-256 Merkle audit of the reasoning database. The site stresses it provides functional agency, not consciousness. It is MIT licensed and starts with npx.

## Features

- **Hierarchical goals:** Decomposes objectives into DAGs with rolled-up progress.
- **Expected utility ranking:** Scores actions against configurable utility weights.
- **Risk assessment:** Rates failure severity, reversibility and blast radius.
- **Belief decay:** Fades belief confidence over time and sweeps stale facts.
- **Reactive replanning:** Rebuilds goals when blockers or events arrive.
- **Fast decision primitives:** Five local tools for labelling, probability, choice and scoring.
- **Audit and traces:** Explainable decision traces and a Merkle audit.

## Good for

- Give an autonomous agent a structured plan with sub-goals
- Rank possible actions under cautious or aggressive profiles
- Check an action's blast radius before dispatching it
- Replan when an unexpected blocker appears
- Explain why an agent chose an action

## Install

### Claude Code (Terminal)

```sh
claude mcp add agent-reasoning-mcp -- npx -y @putervision/agent-reasoning-mcp
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "agent-reasoning-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@putervision/agent-reasoning-mcp"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "agent-reasoning-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@putervision/agent-reasoning-mcp"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "agent-reasoning-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@putervision/agent-reasoning-mcp"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "agent-reasoning-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@putervision/agent-reasoning-mcp"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "agent-reasoning-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@putervision/agent-reasoning-mcp"
      ]
    }
  }
}
```

## Tools

- `set_goal`: Manages goal hierarchy, task DAGs and success criteria.
- `evaluate_situation`: Scores and ranks candidate actions from environment snapshots.
- `replan`: Rebuilds subgoals after blockers or events and aborts stale intentions.
- `assess_risk`: Calculates failure, reversibility and blast-radius risk for actions or plans.
- `query_knowledge`: Searches heuristics, tactics and past decision patterns.
- `set_utility_weights`: Configures and activates utility weight profiles.
- `get_decision_trace`: Returns explainable rationale and latency stats.
- `manage_beliefs`: Maintains beliefs with exponential confidence decay.
- `manage_intentions`: Creates, dispatches and tracks intention directives.
- `manage_reasoning_db`: Provides stats, audit, snapshots and restore.
- `classify`: Fast categorical labelling over state packets.
- `ask_noul`: Typed probabilistic Boolean verification.
- `ask_choice`: Discrete one-of-N choice selection.
- `ask_score`: Bounded numeric utility scoring.
- `gate_intention`: Blast-radius audit gate issuing signed dispatch tokens.

## FAQ

### How do I start?

Run npx @putervision/agent-reasoning-mcp init, then doctor and inspect.

### Does it call an external AI?

The fast layer runs locally with deterministic heuristics and no external API calls.

### How many tools does it have?

15: ten strategic deliberation tools and five fast decision tools.

### What licence?

MIT, per the README badge.

### Is Strategic Agent Reasoning MCP free?

Yes. Strategic Agent Reasoning MCP is open source, so you can use and self-host it for free.

### How do I add Strategic Agent Reasoning MCP to Claude Code?

Run this command in your terminal: claude mcp add agent-reasoning-mcp -- npx -y @putervision/agent-reasoning-mcp

### What are the best alternatives to Strategic Agent Reasoning MCP?

The closest alternatives to Strategic Agent Reasoning MCP are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Concord MCP](https://ai-powered.com/mcp/concord-mcp): Concord MCP is a freemium MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/strategic-agent-reasoning-mcp
