ReasonKit Think MCP

https://docs.reasonkit.sh
Admin
ReasonKit Think is a Model Context Protocol (MCP) server that provides AI coding agents with...
Tokens:14,655
Snippets:246
Trust Score:2.5
Update:5 months ago
Show doc for...
Tokens:
Raw
### Install ReasonKit Think MCP via Crates.io

Source: https://docs.reasonkit.sh/docs/index.html

Use this command to install the ReasonKit Think MCP client if you have Rust 1.95+ and an rmcp-compatible MCP host. This is the recommended installation method.

```bash
cargo install reasonkit-think-mcp
```

--------------------------------

### Install ReasonKit Think with curl

Source: https://docs.reasonkit.sh/

Use this command to install ReasonKit Think. It automatically detects your AI clients and configures MCP.

```shell
curl -fsSL https://get.reasonkit.sh/think | bash
```

--------------------------------

### MCP Server Configuration (`mcp.json`)

Source: https://docs.reasonkit.sh/docs/cursor.html

Example JSON configuration for setting up a custom MCP server named 'reasonkit-think' to run via stdio. Ensure the command path is absolute.

```json
{
  "mcpServers": {
    "reasonkit-think": {
      "type": "stdio",
      "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
      "args": [],
      "env": { "TMPDIR": "/tmp" }
    }
  }
}
```

--------------------------------

### Example Agent Command with Auto Mode

Source: https://docs.reasonkit.sh/

Instructs an AI agent to use Auto mode for reasoning and provide an auditable recommendation. This leverages ReasonKit's structured thinking capabilities.

```bash
"Use Auto mode and return an auditable recommendation."
```

--------------------------------

### Configure MCP Server in Codex Config

Source: https://docs.reasonkit.sh/docs/codex-cli.html

Example TOML configuration for setting up a managed MCP server within the Codex configuration file. Ensure the command path is absolute.

```toml
[mcp_servers.reasonkit-think]
command = "/ABSOLUTE/PATH/TO/reasonkit-think-mcp"
args = []

[mcp_servers.reasonkit-think.env]
TMPDIR = "/tmp"
```

--------------------------------

### Build ReasonKit Think from Source

Source: https://docs.reasonkit.sh/docs/index.html

Follow these steps to build the ReasonKit Think project from its GitHub repository. Ensure you have Rust 1.95+ installed. This method allows for custom builds and local development.

```bash
git clone https://github.com/reasonkit/reasonkit-think.git
cd reasonkit-think
export CARGO_TARGET_DIR="${CARGO_TARGET_DIR:-$PWD/target}"
cargo build --release
```

--------------------------------

### Get All Skeleton-of-Thought Phases

Source: https://docs.reasonkit.sh/docs/skeleton-phases.html

Retrieves all Skeleton-of-Thought phase records for a given deliberation session.

```APIDOC
## GET /websites/reasonkit_sh/skeleton-phases

### Description
Returns all Skeleton-of-Thought phase records for a deliberation.

### Method
GET

### Endpoint
`reasoning://session/{id}/skeleton-phases`

### Parameters
#### Path Parameters
- **id** (string) - Required - The ID of the deliberation session.

### Response
#### Success Response (200)
- **deliberation_id** (string) - The ID of the deliberation.
- **phase_count** (integer) - The total number of phases.
- **phase_ids** (array of strings) - A list of phase IDs.
- **skeleton_phases** (array of objects) - A list of skeleton phase records.
```

--------------------------------

### Get Single Skeleton-of-Thought Phase

Source: https://docs.reasonkit.sh/docs/skeleton-phases.html

Retrieves a single Skeleton-of-Thought phase record by its ID.

```APIDOC
## GET /websites/reasonkit_sh/skeleton-phase/{phase_id}

### Description
Returns one phase record.

### Method
GET

### Endpoint
`reasoning://session/{id}/skeleton-phase/{phase_id}`

### Parameters
#### Path Parameters
- **id** (string) - Required - The ID of the deliberation session.
- **phase_id** (string) - Required - The ID of the specific phase.

### Response
#### Success Response (200)
- **phase_id** (string) - The ID of the phase.
- **source_node_id** (string) - The ID of the source node.
- **state** (string) - The state of the phase.
- **style** (string) - The style of the phase.
- **suitability** (number) - The suitability score of the phase.
- **skeleton_node_ids** (array of strings) - A list of skeleton node IDs.
- **elaboration_node_ids_by_skeleton** (object) - A mapping of skeleton node IDs to elaboration node IDs.
- **prompt_pack** (object) - The prompt pack used for the phase.
- **score_run_ids** (array of strings) - A list of score run IDs.
- **checkpoint_ids** (array of strings) - A list of checkpoint IDs.
- **merge_node_id** (string) - The ID of the merge node, if any.
- **analysis** (object) - Analysis results for the phase.
```

--------------------------------

### Get Route Decision

Source: https://docs.reasonkit.sh/docs/route-decision.html

Retrieves the final route decision, including confidence scores and any identified blockers for a specific session.

```APIDOC
## GET reasoning://session/{id}/route-decision

### Description
Returns final route decision, confidence, and blockers.

### Method
GET

### Endpoint
reasoning://session/{id}/route-decision

### Parameters
#### Path Parameters
- **id** (string) - Required - The unique identifier of the session.
```

--------------------------------

### Run ReasonKit Think Smoke Test

Source: https://docs.reasonkit.sh/docs/index.html

Execute this script after building from source to perform a quick health check of the ReasonKit Think server and its tool surface. This helps verify the installation and basic functionality.

```bash
./scripts/smoke_test.py
```

--------------------------------

### Get Session Graph

Source: https://docs.reasonkit.sh/docs/session-graph.html

Retrieves the full thought graph for a given session ID. The graph includes nodes, typed edges, scores, frontier markers, and metadata.

```APIDOC
## GET reasoning://session/{id}/graph

### Description
Returns the full thought graph for a session, including nodes, typed edges, scores, frontier markers, and metadata.

### Method
GET

### Endpoint
reasoning://session/{id}/graph

### Parameters
#### Path Parameters
- **id** (string) - Required - The unique identifier of the session.
```

--------------------------------

### Get Frontier Nodes

Source: https://docs.reasonkit.sh/docs/frontier.html

Retrieves the active frontier nodes that are eligible for expansion within a given reasoning session. This endpoint is useful for understanding the current state of exploration in a session.

```APIDOC
## GET reasoning://session/{id}/frontier

### Description
Returns active frontier nodes eligible for expansion.

### Method
GET

### Endpoint
reasoning://session/{id}/frontier

### Parameters
#### Path Parameters
- **id** (string) - Required - The unique identifier of the reasoning session.
```

--------------------------------

### Build and Verify MCP Server

Source: https://docs.reasonkit.sh/docs/copilot-vscode.html

Troubleshoot 'spawn ... ENOENT' or connection errors by rebuilding the MCP server binary and verifying its executability and compatibility.

```bash
cd /path/to/reasonkit-think
export CARGO_TARGET_DIR="$PWD/target"
cargo build --release
./scripts/client_compat_check.py
copilot mcp get reasonkit-think
```

--------------------------------

### Verify ReasonKit Think Release Binary Compatibility

Source: https://docs.reasonkit.sh/docs/index.html

Run this script to verify the release binary, protocol handshake, schema compatibility, and local AI coding assistant configurations. It's crucial for ensuring seamless integration with tools like Copilot, Cursor, and Claude.

```bash
./scripts/client_compat_check.py
```

--------------------------------

### Build ReasonKit Think MCP Docker Image

Source: https://docs.reasonkit.sh/docs/index.html

Use this command to build a Docker image for the ReasonKit Think MCP server. This image packages the stdio MCP server, suitable for supervised MCP hosts.

```bash
docker build -t reasonkit-think-mcp .
docker run --rm -i reasonkit-think-mcp
```

--------------------------------

### Configure MCP Server in ~/.claude.json

Source: https://docs.reasonkit.sh/docs/claude-code.html

Add this entry to the `mcpServers` map in your `~/.claude.json` file to register a local stdio server. Ensure the command path is absolute. This configuration is for the 'reasonkit-think' server.

```json
"reasonkit-think": {
  "type": "stdio",
  "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
  "args": [],
  "env": { "TMPDIR": "/tmp" }
}
```

--------------------------------

### Configure MCP Server in OpenCode

Source: https://docs.reasonkit.sh/docs/opencode.html

Add this configuration to your `opencode.json` file to enable and define an MCP server. Ensure the command path is absolute. The `TMPDIR` environment variable can be set for the MCP process.

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "reasonkit-think": {
      "type": "local",
      "command": ["/ABSOLUTE/PATH/TO/reasonkit-think-mcp"],
      "enabled": true,
      "environment": { "TMPDIR": "/tmp" }
    }
  }
}
```

--------------------------------

### start_deliberation

Source: https://docs.reasonkit.sh/docs/start-deliberation.html

Initializes a deliberate reasoning run with specified budgets, strategy, and governance profile.

```APIDOC
## start_deliberation

### Description
Initialize a deliberate reasoning run with budgets, strategy, and governance profile.

### Input
* `session_id` (optional)
* `mode`: `cot | tot | got | reasonkit`
  * accepted user-facing aliases: `quick`, `explore`, `map`, `sketch`, `test`
  * use `run_thinking_mode` for `Auto`; `start_deliberation` only creates a session
* `goal`: string
* `constraints`: string[]
* `profile`: `quick | balanced | deep | paranoid`
* `limits`:
  * `max_depth`
  * `max_nodes`
  * `max_branches_per_expand`
  * `max_verification_questions`

### Output
* `deliberation_id`
* `session_id`
* `mode`
* `limits_applied`
* `status`

### Behavior
* Creates thought graph root node.
* Applies defaults and clamps unsafe limits.
* Returns deterministic IDs and timestamps.
* Keeps wire values compatible while accepting the new Thinking Mode aliases.

### Errors
* Invalid mode/profile
* Limits out of range
* Budget conflict
```

--------------------------------

### Gemini CLI MCP Server Configuration

Source: https://docs.reasonkit.sh/docs/gemini-cli.html

Configure a custom tool as an MCP server in `~/.gemini/settings.json`. Ensure the command path is absolute and specify any necessary environment variables.

```json
"reasonkit-think": {
  "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
  "args": [],
  "env": { "TMPDIR": "/tmp" },
  "trust": true
}
```

--------------------------------

### Run Thinking Mode with Auto

Source: https://docs.reasonkit.sh/docs/index.html

This tool is used to initiate a reasoning process using a specified thinking mode and intent. 'Auto' mode is the default entry point for most work, allowing the system to adaptively choose the best reasoning path.

```json
{
  "tool": "run_thinking_mode",
  "arguments": {
    "mode": "Auto",
    "intent": "Compare the migration options, verify risky claims, and return an auditable recommendation."
  }
}
```

--------------------------------

### Configure VS Code MCP Server

Source: https://docs.reasonkit.sh/docs/copilot-vscode.html

Set up a custom MCP server within VS Code's MCP settings. The configuration is similar to the CLI but uses a different top-level key.

```json
{
  "servers": {
    "reasonkit-think": {
      "type": "stdio",
      "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
      "args": [],
      "env": { "TMPDIR": "/tmp" }
    }
  }
}
```

--------------------------------

### Rebuild and Verify ReasonKit Think Binary

Source: https://docs.reasonkit.sh/docs/index.html

If you encounter connection errors like `ENOENT` or 'failed to connect' when using the ReasonKit Think MCP client, rebuild the release binary and rerun the compatibility check. This ensures the binary is correctly built and configured.

```bash
export CARGO_TARGET_DIR="$PWD/target"
cargo build --release
./scripts/client_compat_check.py
```

--------------------------------

### Configure Copilot CLI MCP Server

Source: https://docs.reasonkit.sh/docs/copilot-vscode.html

Add a custom MCP server to your Copilot CLI configuration. Ensure the command path is absolute and the server is trusted.

```json
{
  "mcpServers": {
    "reasonkit-think": {
      "type": "local",
      "command": "/ABSOLUTE/PATH/TO/reasonkit-think-mcp",
      "args": [],
      "env": { "TMPDIR": "/tmp" },
      "tools": ["*"]
    }
  }
}
```

--------------------------------

### set_verification_policy

Source: https://docs.reasonkit.sh/docs/set-verification-policy.html

Sets or updates the session-level verification policy. This policy will be used for subsequent calls to verify_thoughts, consensus_answer, and run_reasonkit_pipeline.

```APIDOC
## set_verification_policy

### Description
Set or update the session-level verification policy once, then reuse it for subsequent `verify_thoughts`, `consensus_answer`, and `run_reasonkit_pipeline` calls.

### Input

* `deliberation_id`
* `verification_policy`:
  * `min_independent_groups`: integer (1-8)
  * `allow_tier3_for_independence`: boolean
  * `require_tier1_unambiguous_for_critical`: boolean
  * `fail_closed_on_critical_unresolved`: boolean

### Output

* `deliberation_id`
* `verification_policy` (normalized)
* `status` (`updated`)

### Behavior

* Persists policy on the deliberation session.
* Applies normalization and clamps out-of-range values.
* Avoids repeating `policy_override` on each call.

### Errors

* Unknown `deliberation_id`
```

--------------------------------

### Define Custom Reasoning Aliases

Source: https://docs.reasonkit.sh/docs/set-reasoning-aliases.html

Provide a JSON object to define or update trigger phrases for different reasoning modes. The `merge` field controls whether new aliases are added to existing ones or replace them entirely.

```json
{
  "aliases": {
    "auto": ["choose the path"],
    "quick": ["rapid reasoning", "quick logic pass"],
    "explore": ["compare paths"],
    "map": ["dependency synthesis"],
    "sketch": ["outline first"],
    "test": ["evidence gate"],
    "cot": [],
    "tot": [],
    "got": [],
    "verification": [],
    "governance": [],
    "auditability": [],
    "cleanup": []
  },
  "merge": true
}
```
Tokens:
Raw