mcp-sequentialthinking-tools

Name: mcp-sequentialthinking-tools
Author: spences10

3.5kinstalls/mo · npm582stars87forks

How to install mcp-sequentialthinking-tools?

spences10/mcp-sequentialthinking-tools

$git clone https://github.com/spences10/mcp-sequentialthinking-tools

Installs into the current project.

Summary

A lightweight MCP server for recording sequential reasoning steps. It is a scratchpad with history, branching, revision metadata, and optional validation for model-authored tool plans.

Files · 1

View on GitHub

README.md

1	# mcp-sequentialthinking-tools
2
3	[![built with vite+](https://img.shields.io/badge/built%20with-Vite+-646CFF?logo=vite&logoColor=white)](https://viteplus.dev)
4	[![tested with vitest](https://img.shields.io/badge/tested%20with-Vitest-6E9F18?logo=vitest&logoColor=white)](https://vitest.dev)
5
6	A lightweight MCP server for recording sequential reasoning steps. It
7	is a scratchpad with history, branching, revision metadata, and
8	optional validation for model-authored tool plans.
9
10	It does not discover your other MCP tools and it does not
11	choose tools for the model. If you pass `available_tools` and
12	`recommended_tools`, the server validates that the recommended names
13	exist and stores the step.
14
15	<a href="https://glama.ai/mcp/servers/zl990kfusy">
16	<img width="380" height="200" src="https://glama.ai/mcp/servers/zl990kfusy/badge" />
17	</a>
18
19	## Why use it?
20
21	Use this when a task benefits from explicit, inspectable reasoning:
22
23	- breaking a messy problem into steps;
24	- revising or branching a plan;
25	- keeping a small reasoning history by session;
26	- validating tool-plan names against a supplied tool list;
27	- clearing or inspecting reasoning history during a long agent run.
28
29	Do not use it for trivial requests. It adds overhead.
30
31	## Tools
32
33	### `sequentialthinking_tools`
34
35	Records one thought.
36
37	Required parameters:
38
39	- `thought` — current reasoning step
40	- `thought_number` — current step number
41	- `total_thoughts` — current estimate; automatically raised if lower
42	than `thought_number`
43	- `next_thought_needed` — whether another thought is needed
44
45	Optional parameters:
46
47	- `session_id` — history bucket; defaults to `default`
48	- `is_revision`, `revises_thought`
49	- `branch_from_thought`, `branch_id`
50	- `needs_more_thoughts`
51	- `available_tools` — array of tool names or `{ name, description }`
52	objects
53	- `recommended_tools` — model-authored recommendations to
54	validate/store
55	- `remaining_steps` — short list of upcoming steps
56
57	Example:
58
59	```json
60	{
61	"session_id": "svelte-debug",
62	"thought": "First inspect the route files, then run the failing check.",
63	"thought_number": 1,
64	"total_thoughts": 3,
65	"next_thought_needed": true,
66	"available_tools": ["read", "bash"],
67	"recommended_tools": [
68	{
69	"tool_name": "read",
70	"confidence": 0.9,
71	"rationale": "Need to inspect the relevant files before editing.",
72	"priority": 1
73	}
74	]
75	}
76	```
77
78	If `recommended_tools` contains a name not present in
79	`available_tools`, the call returns `isError: true` and does not store
80	the thought.
81
82	## Security posture
83
84	The server treats thought text, tool descriptions, rationales, and
85	remaining-step text as untrusted input. Prompt-injection-like text is
86	scanned and redacted before it is stored or returned in history. Calls
87	with redactions include `security_warnings` showing which fields
88	matched.
89
90	This is defensive filtering, not a guarantee that arbitrary
91	adversarial text is safe. Do not put secrets in thoughts or tool
92	descriptions.
93
94	### `get_thinking_history`
95
96	Returns stored thoughts for a session.
97
98	Parameters:
99
100	- `session_id` — defaults to `default`
101	- `branch_id` — optional branch filter
102	- `limit` — max records to return; default `50`, max `500`
103
104	### `clear_thinking_history`
105
106	Clears one session or every session.
107
108	Parameters:
109
110	- `session_id` — defaults to `default`
111	- `all_sessions` — clear all history buckets
112
113	## Prompt
114
115	### `sequential-thinking-guidance`
116
117	A short prompt that tells the model how to use this server honestly:
118	as a scratchpad and validator, not as an external reasoning engine.
119
120	## Configuration
121
122	### Claude Desktop / compatible MCP clients
123
124	```json
125	{
126	"mcpServers": {
127	"mcp-sequentialthinking-tools": {
128	"command": "npx",
129	"args": ["-y", "mcp-sequentialthinking-tools"],
130	"env": {
131	"MAX_HISTORY_SIZE": "1000"
132	}
133	}
134	}
135	}
136	```
137
138	`MAX_HISTORY_SIZE` is per session and defaults to `1000`.
139
140	The server uses `tmcp` and includes a small stdio transport that
141	accepts both standard `Content-Length` framed MCP messages and
142	newline-delimited JSON used by older `tmcp` tooling.
143
144	## Dev

People who liked this also viewed

6 picks

Type

mcp-sequentialthinking-tools

3.5kinstalls/mo · npm582stars87forks

View on GitHub

How to install mcp-sequentialthinking-tools?

spences10/mcp-sequentialthinking-tools

$git clone https://github.com/spences10/mcp-sequentialthinking-tools

Installs into the current project.

Summary

A lightweight MCP server for recording sequential reasoning steps. It is a scratchpad with history, branching, revision metadata, and optional validation for model-authored tool plans.

Files · 1

View on GitHub

README.md

1	# mcp-sequentialthinking-tools
2
3	[![built with vite+](https://img.shields.io/badge/built%20with-Vite+-646CFF?logo=vite&logoColor=white)](https://viteplus.dev)
4	[![tested with vitest](https://img.shields.io/badge/tested%20with-Vitest-6E9F18?logo=vitest&logoColor=white)](https://vitest.dev)
5
6	A lightweight MCP server for recording sequential reasoning steps. It
7	is a scratchpad with history, branching, revision metadata, and
8	optional validation for model-authored tool plans.
9
10	It does not discover your other MCP tools and it does not
11	choose tools for the model. If you pass `available_tools` and
12	`recommended_tools`, the server validates that the recommended names
13	exist and stores the step.
14
15	<a href="https://glama.ai/mcp/servers/zl990kfusy">
16	<img width="380" height="200" src="https://glama.ai/mcp/servers/zl990kfusy/badge" />
17	</a>
18
19	## Why use it?
20
21	Use this when a task benefits from explicit, inspectable reasoning:
22
23	- breaking a messy problem into steps;
24	- revising or branching a plan;
25	- keeping a small reasoning history by session;
26	- validating tool-plan names against a supplied tool list;
27	- clearing or inspecting reasoning history during a long agent run.
28
29	Do not use it for trivial requests. It adds overhead.
30
31	## Tools
32
33	### `sequentialthinking_tools`
34
35	Records one thought.
36
37	Required parameters:
38
39	- `thought` — current reasoning step
40	- `thought_number` — current step number
41	- `total_thoughts` — current estimate; automatically raised if lower
42	than `thought_number`
43	- `next_thought_needed` — whether another thought is needed
44
45	Optional parameters:
46
47	- `session_id` — history bucket; defaults to `default`
48	- `is_revision`, `revises_thought`
49	- `branch_from_thought`, `branch_id`
50	- `needs_more_thoughts`
51	- `available_tools` — array of tool names or `{ name, description }`
52	objects
53	- `recommended_tools` — model-authored recommendations to
54	validate/store
55	- `remaining_steps` — short list of upcoming steps
56
57	Example:
58
59	```json
60	{
61	"session_id": "svelte-debug",
62	"thought": "First inspect the route files, then run the failing check.",
63	"thought_number": 1,
64	"total_thoughts": 3,
65	"next_thought_needed": true,
66	"available_tools": ["read", "bash"],
67	"recommended_tools": [
68	{
69	"tool_name": "read",
70	"confidence": 0.9,
71	"rationale": "Need to inspect the relevant files before editing.",
72	"priority": 1
73	}
74	]
75	}
76	```
77
78	If `recommended_tools` contains a name not present in
79	`available_tools`, the call returns `isError: true` and does not store
80	the thought.
81
82	## Security posture
83
84	The server treats thought text, tool descriptions, rationales, and
85	remaining-step text as untrusted input. Prompt-injection-like text is
86	scanned and redacted before it is stored or returned in history. Calls
87	with redactions include `security_warnings` showing which fields
88	matched.
89
90	This is defensive filtering, not a guarantee that arbitrary
91	adversarial text is safe. Do not put secrets in thoughts or tool
92	descriptions.
93
94	### `get_thinking_history`
95
96	Returns stored thoughts for a session.
97
98	Parameters:
99
100	- `session_id` — defaults to `default`
101	- `branch_id` — optional branch filter
102	- `limit` — max records to return; default `50`, max `500`
103
104	### `clear_thinking_history`
105
106	Clears one session or every session.
107
108	Parameters:
109
110	- `session_id` — defaults to `default`
111	- `all_sessions` — clear all history buckets
112
113	## Prompt
114
115	### `sequential-thinking-guidance`
116
117	A short prompt that tells the model how to use this server honestly:
118	as a scratchpad and validator, not as an external reasoning engine.
119
120	## Configuration
121
122	### Claude Desktop / compatible MCP clients
123
124	```json
125	{
126	"mcpServers": {
127	"mcp-sequentialthinking-tools": {
128	"command": "npx",
129	"args": ["-y", "mcp-sequentialthinking-tools"],
130	"env": {
131	"MAX_HISTORY_SIZE": "1000"
132	}
133	}
134	}
135	}
136	```
137
138	`MAX_HISTORY_SIZE` is per session and defaults to `1000`.
139
140	The server uses `tmcp` and includes a small stdio transport that
141	accepts both standard `Content-Length` framed MCP messages and
142	newline-delimited JSON used by older `tmcp` tooling.
143
144	## Dev

People who liked this also viewed

6 picks

Type