paput-mcp 2.5.0 → 2.6.1

package/docs/privacy-policy.md

@@ -9,7 +9,7 @@ PaPut MCP Server connects AI assistants and MCP clients to PaPut through the Mod
 Depending on the tools you use, the server may process:
 
 - PaPut memos, notes, categories, skill sheet data, and related project metadata.
-- OAuth access tokens or API keys used to authenticate requests to PaPut.
+- OAuth access tokens used to authenticate requests to PaPut.
 - MCP request and response metadata needed to execute tool calls.
 - Local Claude or Codex session metadata and transcripts when you explicitly use local knowledge capture tools.
 - Local pending knowledge candidates and cache data stored on your device.
@@ -28,15 +28,15 @@ Data is used only to provide the MCP tools you invoke, including:
 
 Remote MCP connections use OAuth. PaPut issues tokens after you sign in and approve the requested scopes. The MCP server receives bearer tokens on requests and forwards them to the PaPut API. The MCP server does not intentionally persist OAuth access tokens.
 
-Local stdio usage may use a PaPut API key provided through environment variables. You are responsible for keeping API keys and local configuration files private.
+Local stdio usage uses OAuth tokens created by `paput-mcp login`. You are responsible for keeping local token and configuration files private.
 
 ## Local Data
 
-When local knowledge capture features are used, PaPut MCP may store cache files under `~/.paput` or a configured cache directory. This local data can include synced memo summaries, pending knowledge candidates, processed session markers, and session metadata. It remains on your device unless you choose to save a pending candidate to PaPut.
+When local knowledge capture features are used, PaPut MCP may store cache files under `~/.paput` or a configured cache directory. This local data can include synced memo summaries, pending knowledge candidates, processed session markers, session metadata, and local OAuth tokens created by `paput-mcp login`. The default OAuth token directory is created with `0700` permissions and the token file is written with `0600` permissions. It remains on your device unless you choose to save a pending candidate to PaPut or remove the local token cache with `paput-mcp logout`.
 
 ## Logging
 
-Hosted infrastructure and MCP clients may record operational logs such as request timestamps, status codes, errors, and connection metadata. Logs are used for reliability, troubleshooting, abuse prevention, and security. The server is designed not to log API keys, OAuth tokens, or full private content intentionally.
+Hosted infrastructure and MCP clients may record operational logs such as request timestamps, status codes, errors, and connection metadata. Logs are used for reliability, troubleshooting, abuse prevention, and security. The server is designed not to log OAuth tokens or full private content intentionally.
 
 ## Sharing
 
@@ -49,6 +49,7 @@ You can:
 - Revoke OAuth access from PaPut account settings when available.
 - Remove local MCP configuration from your client.
 - Delete local cache data stored under `~/.paput` or your configured cache directory.
+- Run `paput-mcp logout` to revoke and remove the local OAuth token cache.
 - Delete or update PaPut content using PaPut or authorized MCP tools.
 
 ## Security

package/docs/tools.md

@@ -2,6 +2,10 @@
 
 PaPut MCP is a data-only MCP server. It exposes tools and resources for PaPut data. It does not provide UI widgets.
 
+Remote HTTP mode exposes API-backed PaPut tools only. Local cache, pending
+candidate, and session transcript tools are available in local CLI mode because
+they read or write files on the user's device.
+
 ## Confirmation Policy
 
 Clients and assistants should follow these rules:
@@ -25,6 +29,10 @@ Clients and assistants should follow these rules:
 | `paput_update_memo`    | Destructive/write | Update an existing memo title, body, visibility, categories, or linked projects. |
 | `paput_get_categories` | Read-only         | List categories before assigning categories or checking duplicates.              |
 
+`paput_create_memo` and `paput_update_memo` can link projects through explicit
+`projects`, a `project_match` input, or local `PAPUT_PROJECT_MATCH` in local CLI
+mode.
+
 ## Note Tools
 
 | Tool                 | Safety            | Use case                                               |
@@ -47,6 +55,9 @@ Clients and assistants should follow these rules:
 
 ## Knowledge Capture And Local Cache Tools
 
+These tools are local CLI mode only. They are not listed by the remote HTTP MCP
+server.
+
 | Tool                              | Safety                       | Use case                                                                      |
 | --------------------------------- | ---------------------------- | ----------------------------------------------------------------------------- |
 | `paput_cache_status`              | Read-only                    | Inspect local cache, pending candidates, processed sessions, and sync status. |

package/docs/usage-examples.md

@@ -2,7 +2,22 @@
 
 These examples show typical ways to use PaPut MCP from an AI assistant.
 
-## 1. Search Existing Memos Before Creating Knowledge
+## 1. Create A Memo From A Remote MCP Client
+
+Prompt:
+
+```text
+Create a private PaPut memo summarizing this conversation.
+```
+
+Expected tool flow:
+
+1. `paput_create_memo`
+2. `paput_search_memo` or `paput_get_memo` when the user wants to verify the result
+
+Use case: quickly save useful content from Claude, ChatGPT, Codex, or another remote MCP client.
+
+## 2. Search Existing Memos Before Creating Knowledge
 
 Prompt:
 
@@ -18,7 +33,10 @@ Expected tool flow:
 
 Use case: avoid duplicate long-term knowledge while preserving useful decisions from engineering work.
 
-## 2. Capture Reusable Knowledge From A Codex Session
+Local mode note: `paput_add_knowledge_candidates` and pending review tools are
+local CLI mode tools.
+
+## 3. Capture Reusable Knowledge From A Codex Session
 
 Prompt:
 
@@ -34,7 +52,10 @@ Expected tool flow:
 
 Use case: turn completed development work into reviewable knowledge without immediately publishing it to PaPut.
 
-## 3. Save An Approved Pending Candidate
+Local mode note: session scanning and transcript reading are local CLI mode
+tools.
+
+## 4. Save An Approved Pending Candidate
 
 Prompt:
 
@@ -50,7 +71,9 @@ Expected tool flow:
 
 Use case: keep the user in control before creating a permanent PaPut memo.
 
-## 4. Update A Skill Sheet Project
+Local mode note: pending candidate tools are local CLI mode tools.
+
+## 5. Update A Skill Sheet Project
 
 Prompt:
 
@@ -66,7 +89,7 @@ Expected tool flow:
 
 Use case: maintain an accurate skill sheet after project milestones.
 
-## 5. Organize Memos Into A Note
+## 6. Organize Memos Into A Note
 
 Prompt:
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "paput-mcp",
-  "version": "2.5.0",
+  "version": "2.6.1",
   "description": "PaPut MCP Server",
   "main": "dist/index.js",
   "type": "module",