@fileverse/api 0.0.10 → 0.0.11

package/public/llm.txt

@@ -1,621 +1,347 @@
 # Fileverse API - AI Agent Guide
 
-> Fileverse API is a document management system by Fileverse that stores and syncs documents (ddocs) between a local database and blockchain.
+> Fileverse API lets you create, manage, and publish documents to blockchain. This guide helps you (the AI agent) get connected and start working with documents.
 
-**Skill Name**: fileverse-api
+## Step 1: Connect via MCP (Recommended)
 
-## How to use this document
+The fastest way to start is by adding the Fileverse MCP server to your client. Ask the user to add the following configuration to their AI tool:
 
-Fetch this via GET {SERVER_URL}/llm.txt. Use it to discover the server URL, decide API vs CLI mode, then follow the relevant section below.
+### Claude Code
 
-## Integration checklist
+Run this command, replacing the placeholder values:
 
-1. Get server URL (from the URL you used to fetch this doc, minus /llm.txt) and apiKey (see API Key Configuration).
-2. Try GET {SERVER_URL}/ping. If you get 200 and {"reply":"pong"}, use API mode; otherwise use CLI mode.
-3. In API mode: every request must include apiKey as a query parameter; the path is complete before any query string.
-4. For create/update: extract ddocId from response.data.ddocId (not from the top level of the response).
-5. After create or update, poll GET /api/ddocs/{ddocId}?apiKey=... until syncStatus is "synced" or "failed", then show the link if synced.
-
-## Server URL
-
-If you fetched this document from a URL (e.g., http://localhost:8001/llm.txt):
-- Your server URL is that URL without the /llm.txt suffix
-- Example: http://localhost:8001/llm.txt → Server URL is http://localhost:8001
-
-If this document was provided another way, ask the user for their server URL.
-
-## API Key Configuration
-
-Before starting, check for stored credentials:
-
-1. Look for config file at: ~/.fileverseapirc
-2. If found with valid apiKey, use it automatically
-3. If not found or missing apiKey:
-   - Ask user for their Fileverse API key
-   - Offer to save it: "Would you like me to save this API key for future sessions?"
-   - If yes, create/update ~/.fileverseapirc with JSON: {"apiKey": "...", "serverUrl": "..."}
-
-## Connection Strategy
-
-After getting the server URL:
-
-1. **Try API first**: Attempt GET {SERVER_URL}/ping
-2. **If successful** (response: {"reply":"pong"}) → Use API Mode
-3. **If connection fails** (timeout, refused, unreachable) → Switch to CLI Mode
-
-In API mode, every request must include apiKey as a query parameter; the path is complete before any query string.
-
-When switching to CLI mode, inform the user:
-"I can't reach your server directly (localhost isn't accessible from here). I'll help you using CLI commands instead - just run them and share the output with me."
-
----
-
-# Quick reference
-
-## All API endpoints
-
-| Method | Path | Auth | Key params/body |
-|--------|------|------|-----------------|
-| GET | /ping | No | - |
-| GET | /api/ddocs | apiKey | limit, skip |
-| GET | /api/ddocs/{ddocId} | apiKey | - |
-| POST | /api/ddocs | apiKey | JSON: {title, content} or form-data: file |
-| PUT | /api/ddocs/{ddocId} | apiKey | JSON: {title?, content?} or form-data: file |
-| DELETE | /api/ddocs/{ddocId} | apiKey | - |
-| GET | /api/folders | apiKey | limit, skip |
-| POST | /api/folders | apiKey | body: see Create Folder |
-| GET | /api/folders/{folderRef}/{folderId} | apiKey | - |
-| GET | /api/search | apiKey | q (required), limit, skip |
-| GET | /api/events/failed | apiKey | - |
-| POST | /api/events/retry-failed | apiKey | - |
-| POST | /api/events/{id}/retry | apiKey | - |
-
-## All ddctl commands
-
-| Command | Purpose | Key options |
-|---------|---------|-------------|
-| ddctl list | List ddocs | -l/--limit, -s/--skip |
-| ddctl get <ddocId> | Get document metadata | - |
-| ddctl view <ddocId> | View content preview | -n/--lines (default 10) |
-| ddctl create <filepath> | Create from file | title = filename; file cannot be empty |
-| ddctl update <ddocId> | Update ddoc | -f/--file <path> or opens editor |
-| ddctl download <ddocId> | Download to file | -o/--output <filename> |
-| ddctl delete <ddocIds...> | Delete one or more | space-separated ids |
-| ddctl events list-failed | List failed worker events | - |
-| ddctl events retry <eventId> | Retry one failed event | - |
-| ddctl events retry-all | Retry all failed events | - |
-
----
-
-# MCP Integration (Recommended for AI Agents)
-
-The Fileverse API MCP server is the easiest way for AI agents to interact with Fileverse API. It wraps all API calls into native MCP tools with built-in sync polling.
-
-## Setup
-
-Install Fileverse API globally, then configure your MCP client:
-
-### Claude Code / Cursor
+```bash
+claude mcp add fileverse-api -- npx -y @anthropic-ai/mcp-proxy@latest --endpoint "YOUR_SERVER_URL/mcp" --header "x-api-key: YOUR_API_KEY"
+```
 
-Add to your MCP config (e.g., `.claude/mcp.json` or Cursor MCP settings):
+Or add to `.claude/mcp.json`:
 
 ```json
 {
   "mcpServers": {
     "fileverse-api": {
-      "command": "fileverse-api-mcp",
-      "env": {
-        "FILEVERSE_API_KEY": "your-api-key",
-        "FILEVERSE_SERVER_URL": "http://localhost:8001"
-      }
+      "command": "npx",
+      "args": ["-y", "@anthropic-ai/mcp-proxy@latest", "--endpoint", "YOUR_SERVER_URL/mcp", "--header", "x-api-key: YOUR_API_KEY"]
     }
   }
 }
 ```
 
-The MCP server can also read credentials from `~/.fileverseapirc` if env vars are not set.
-
-## Available MCP Tools
-
-| Tool | Description |
-|------|-------------|
-| fileverse_list_documents | List documents with pagination |
-| fileverse_get_document | Get a single document by ddocId |
-| fileverse_create_document | Create document and wait for sync (returns link) |
-| fileverse_update_document | Update document and wait for sync (returns link) |
-| fileverse_delete_document | Delete a document |
-| fileverse_search_documents | Search documents by text query |
-| fileverse_get_sync_status | Check sync status and link of a document |
-| fileverse_retry_failed_events | Retry all failed blockchain sync events |
-
-Create and update tools automatically poll until blockchain sync completes, so the agent does not need to implement polling logic.
-
----
-
-# Response Schemas
-
-## Document
-
-Fields returned by GET /api/ddocs/{ddocId} and in list/search results:
-
-```
-{
-  ddocId: string,            // Unique document identifier
-  title: string,             // Document title
-  content: string,           // Document content (text/markdown)
-  syncStatus: "pending" | "synced" | "failed",
-  link: string | null,       // Public URL (only when syncStatus is "synced")
-  localVersion: number,      // Increments on each local edit
-  onchainVersion: number,    // Version published to blockchain
-  isDeleted: number,         // 0 = active, 1 = deleted
-  onChainFileId: number | null,
-  portalAddress: string,
-  createdAt: string,         // ISO timestamp
-  updatedAt: string          // ISO timestamp
-}
-```
+### Cursor
 
-## Document List
+Add to Cursor MCP settings (`.cursor/mcp.json`):
 
-Returned by GET /api/ddocs:
-
-```
+```json
 {
-  ddocs: Document[],
-  total: number,
-  hasNext: boolean
+  "mcpServers": {
+    "fileverse-api": {
+      "command": "npx",
+      "args": ["-y", "@anthropic-ai/mcp-proxy@latest", "--endpoint", "YOUR_SERVER_URL/mcp", "--header", "x-api-key: YOUR_API_KEY"]
+    }
+  }
 }
 ```
 
-## Search Result
+### Windsurf
 
-Returned by GET /api/search:
+Add to `~/.codeium/windsurf/mcp_config.json`:
 
-```
+```json
 {
-  nodes: Document[],         // Note: "nodes" not "ddocs"
-  total: number,
-  hasNext: boolean
+  "mcpServers": {
+    "fileverse-api": {
+      "command": "npx",
+      "args": ["-y", "@anthropic-ai/mcp-proxy@latest", "--endpoint", "YOUR_SERVER_URL/mcp", "--header", "x-api-key: YOUR_API_KEY"]
+    }
+  }
 }
 ```
 
-## Create/Update Response
+### ChatGPT (Desktop)
 
-Returned by POST /api/ddocs and PUT /api/ddocs/{ddocId}:
+Go to Settings > MCP Servers > Add Server, then enter:
 
-```
-{
-  message: string,
-  data: Document              // Extract ddocId from data.ddocId
-}
-```
+- **Name**: fileverse-api
+- **Command**: `npx -y @anthropic-ai/mcp-proxy@latest --endpoint "YOUR_SERVER_URL/mcp" --header "x-api-key: YOUR_API_KEY"`
 
-## Event
+### Any MCP-compatible client
 
-Returned by GET /api/events/failed:
+The Fileverse API exposes an MCP endpoint at `{SERVER_URL}/mcp`. Connect using any MCP client that supports the Streamable HTTP transport. Pass the API key via the `x-api-key` header.
 
-```
-{
-  _id: string,
-  type: "create" | "update" | "delete",
-  timestamp: number,
-  fileId: string,
-  portalAddress: string,
-  status: "pending" | "processing" | "processed" | "failed",
-  retryCount: number,
-  lastError: string | null,
-  userOpHash: string | null
-}
-```
+### Replacing placeholders
 
----
+- **YOUR_SERVER_URL**: The base URL of the Fileverse API server (e.g., `http://localhost:8001`). If the user shared this document via a URL, strip `/llm.txt` to get the server URL.
+- **YOUR_API_KEY**: The user's Fileverse API key. Check `~/.fileverseapirc` for stored credentials, or ask the user to provide it.
 
-# CLI MODE (Fallback when server unreachable)
+### After adding the MCP server
 
-Use this when you cannot reach the user's server (localhost, firewall, etc.).
-You generate commands, user runs them and shares output with you.
+Restart or reload the AI tool. You will then have access to these tools:
 
-## Installation and invocation
+| Tool | Description |
+|------|-------------|
+| `fileverse_list_documents` | List documents with pagination |
+| `fileverse_get_document` | Get a single document by ddocId |
+| `fileverse_create_document` | Create document and wait for blockchain sync (returns link) |
+| `fileverse_update_document` | Update document and wait for blockchain sync (returns link) |
+| `fileverse_delete_document` | Delete a document |
+| `fileverse_search_documents` | Search documents by text query |
+| `fileverse_get_sync_status` | Check sync status and blockchain link of a document |
+| `fileverse_retry_failed_events` | Retry all failed blockchain sync events |
 
-To set up Fileverse API and run the server:
+Create and update tools automatically poll until blockchain sync completes, so you do not need to implement polling logic. Once synced, the response includes a public `link` to the document.
 
-```bash
-npm install -g @fileverse/api
-npx @fileverse/api --apiKey <key> --rpcUrl <url>
-```
+---
 
-The package exposes the ddctl binary. After npm install -g @fileverse/api, run ddctl <command>. Without global install: npx -p @fileverse/api ddctl <command>.
+## Step 2 (Alternative): Use the REST API directly
 
-## CLI Commands Reference
+If MCP is not available in your environment, you can call the REST API directly.
 
-### List Documents
-```bash
-ddctl list
-ddctl list -l 10 -s 20
-ddctl list --limit 10 --skip 20
-```
+### Server URL
 
-### Get Document Details
-```bash
-ddctl get <ddocId>
-```
-Shows metadata including the link (if synced to blockchain).
+If this document was fetched from a URL (e.g., `http://localhost:8001/llm.txt`), the server URL is that URL without `/llm.txt`.
 
-### View Document Content
-```bash
-ddctl view <ddocId>
-ddctl view <ddocId> -n 20
-ddctl view <ddocId> --lines 20
-```
-Default preview is 10 lines.
+### Authentication
 
-### Create Document
-```bash
-ddctl create <filepath>
-```
-Creates document from file. Title is derived from filename. File content cannot be empty.
-
-### Update Document
-```bash
-ddctl update <ddocId> -f <filepath>
-ddctl update <ddocId> --file <filepath>
-ddctl update <ddocId>
-```
-Without --file opens in editor (uses EDITOR env or vi).
+Every API request requires an `apiKey` query parameter appended to the URL:
 
-### Download Document
-```bash
-ddctl download <ddocId>
-ddctl download <ddocId> -o myfile.md
-ddctl download <ddocId> --output myfile.md
 ```
-Output supports markdown.
-
-### Delete Document
-```bash
-ddctl delete <ddocId>
-ddctl delete <id1> <id2> <id3>
+{SERVER_URL}{PATH}?apiKey={API_KEY}
 ```
-One or more space-separated ddoc IDs.
 
-### Worker events
-```bash
-ddctl events list-failed
-ddctl events retry <eventId>
-ddctl events retry-all
-```
-If sync is stuck, run ddctl events list-failed then ddctl events retry <id> or ddctl events retry-all.
+Check `~/.fileverseapirc` for stored credentials (JSON with `apiKey` and `serverUrl` fields). If not found, ask the user.
 
-## CLI Workflow Examples
+### Verify connectivity
 
-### Create and Check Sync Status
-```bash
-ddctl create mydocument.md
-ddctl get <ddocId>
 ```
-Repeat get until syncStatus is "synced"; then the link field shows the public URL.
-
-### Search and Update
-```bash
-ddctl list
-ddctl view <ddocId>
-ddctl update <ddocId> --file updated.md
-ddctl get <ddocId>
+GET {SERVER_URL}/ping
 ```
 
----
-
-# API MODE (Primary - when server is reachable)
-
-Use this when GET /ping succeeds. You make HTTP requests directly to the server.
-
-## Authentication & URL Construction
-
-CRITICAL: The apiKey is a query parameter that goes at the END of the URL, after the complete path.
-
-URL Structure: {SERVER_URL}{PATH}?apiKey={API_KEY}&{OTHER_PARAMS}
-
-Step-by-step URL construction:
-1. Start with the server URL: http://localhost:8001
-2. Add the complete path: /api/ddocs/abc123
-3. Add ? and the apiKey: ?apiKey=your-key
-4. Add any other params with &: &limit=10
-
-CORRECT examples:
-- GET http://localhost:8001/api/ddocs?apiKey=xxx
-- GET http://localhost:8001/api/ddocs/abc123?apiKey=xxx
-- PUT http://localhost:8001/api/ddocs/abc123?apiKey=xxx
-- GET http://localhost:8001/api/ddocs?apiKey=xxx&limit=10&skip=0
-
-WRONG (do NOT do this):
-- GET http://localhost:8001/api/ddocs?apiKey=xxx/abc123  ← WRONG: path after query param
-- GET http://localhost:8001/api?apiKey=xxx/ddocs/abc123  ← WRONG: path after query param
-
-The query string (?apiKey=...) must ALWAYS come after the full path is complete.
-
-## Common mistakes
-
-- Putting path segments after ?apiKey=... (wrong).
-- Forgetting that create/update return data.ddocId, not top-level ddocId; always use response.data.ddocId.
-- Not polling after create/update until syncStatus is "synced" or "failed".
-
-## Important: Sync Status Polling
-
-When you create or update a document, it syncs to blockchain asynchronously:
-
-- **pending**: Saved locally, blockchain sync in progress
-- **synced**: Published to blockchain (link field available)
-- **failed**: Sync failed
-
-After CREATE or UPDATE, poll until syncStatus is "synced" or "failed":
-
-1. POST /api/ddocs or PUT /api/ddocs/{id}
-2. Extract ddocId from response.data.ddocId (not from the top level)
-3. Poll: GET /api/ddocs/{ddocId}?apiKey={API_KEY}
-4. If "pending", wait 5 seconds and poll again
-5. If "synced": extract the "link" field and show as clickable: [View Document](link)
-6. If "failed", inform user of sync failure
-
-Typical sync time: 5-30 seconds.
-
-## API Reference
-
-Base: {SERVER_URL}
-
-### Health Check (No Auth)
-GET /ping
-Response: {"reply":"pong"}
-
-### List Documents
-GET /api/ddocs?apiKey={API_KEY}&limit=10&skip=0
-
-Response:
-{
-  "ddocs": [...],
-  "total": 100,
-  "hasNext": true
-}
-
-### Get Document
-GET /api/ddocs/{ddocId}?apiKey={API_KEY}
-
-Response:
-{
-  "ddocId": "...",
-  "title": "...",
-  "content": "...",
-  "syncStatus": "pending|synced|failed",
-  "link": "https://...",
-  ...
-}
-(link available when synced)
-
-### Create Document (JSON - recommended for AI agents)
-POST /api/ddocs?apiKey={API_KEY}
-Content-Type: application/json
-
-{ "title": "Document Title", "content": "Document content here..." }
+Expected response: `{"reply":"pong"}`
 
-Response: 201 with { "message": "...", "data": { "ddocId": "...", "title": "...", "syncStatus": "...", ... } }
-Extract ddocId from response.data.ddocId. Then poll GET /api/ddocs/{ddocId} until syncStatus is "synced", then show the link.
+### API Endpoints
 
-### Create Document (Multipart - for file uploads)
-POST /api/ddocs?apiKey={API_KEY}
-Content-Type: multipart/form-data
+| Method | Path | Description |
+|--------|------|-------------|
+| GET | /ping | Health check (no auth) |
+| GET | /api/ddocs?apiKey=...&limit=10&skip=0 | List documents |
+| GET | /api/ddocs/{ddocId}?apiKey=... | Get one document |
+| POST | /api/ddocs?apiKey=... | Create document (JSON body: `{title, content}`) |
+| PUT | /api/ddocs/{ddocId}?apiKey=... | Update document (JSON body: `{title?, content?}`) |
+| DELETE | /api/ddocs/{ddocId}?apiKey=... | Delete document |
+| GET | /api/search?apiKey=...&q={query} | Search documents |
+| GET | /api/events/failed?apiKey=... | List failed sync events |
+| POST | /api/events/retry-failed?apiKey=... | Retry all failed events |
+| POST | /api/events/{id}/retry?apiKey=... | Retry one failed event |
 
-Send a single form field "file" with the document file. Title is derived from the filename. Max file size 10MB.
+### Sync Status Polling (Required for REST API)
 
-Response: same as JSON create above.
+When using the REST API (not MCP), you must poll after create/update:
 
-### Update Document (JSON - recommended for AI agents)
-PUT /api/ddocs/{ddocId}?apiKey={API_KEY}
-Content-Type: application/json
+1. Send POST or PUT request
+2. Extract `ddocId` from `response.data.ddocId` (not from the top level)
+3. Poll `GET /api/ddocs/{ddocId}?apiKey=...` every 5 seconds
+4. When `syncStatus` is `"synced"`: show the `link` field to the user
+5. When `syncStatus` is `"failed"`: inform the user and suggest retrying
 
-{ "title": "New Title", "content": "Updated content..." }
+### URL Construction (Common Mistake)
 
-Both fields are optional; send only what you want to change.
+The `apiKey` query parameter must come AFTER the complete path:
 
-Response: 200 with { "message": "...", "data": { ... } }
-Poll until synced, then show the link.
-
-### Update Document (Multipart - for file uploads)
-PUT /api/ddocs/{ddocId}?apiKey={API_KEY}
-Content-Type: multipart/form-data
+```
+CORRECT: GET http://localhost:8001/api/ddocs/abc123?apiKey=xxx
+WRONG:   GET http://localhost:8001/api/ddocs?apiKey=xxx/abc123
+```
 
-Send a single form field "file" with the updated document file. Title is derived from the filename. Max file size 10MB.
+---
 
-Response: same as JSON update above.
+## Step 3 (Fallback): CLI Mode
 
-### Delete Document
-DELETE /api/ddocs/{ddocId}?apiKey={API_KEY}
+If you cannot reach the server (e.g., localhost is not accessible), generate CLI commands for the user to run.
 
-### List Folders
-GET /api/folders?apiKey={API_KEY}&limit=10&skip=0
+### Installation
 
-Response:
-{
-  "folders": [...],
-  "total": 100,
-  "hasNext": true
-}
+```bash
+npm install -g @fileverse/api
+```
 
-### Create Folder
-POST /api/folders?apiKey={API_KEY}
-Content-Type: application/json
+### Commands
+
+| Command | Purpose |
+|---------|---------|
+| `ddctl list` | List documents (`-l`/`--limit`, `-s`/`--skip`) |
+| `ddctl get <ddocId>` | Get document metadata |
+| `ddctl view <ddocId>` | Preview content (`-n`/`--lines`, default 10) |
+| `ddctl create <filepath>` | Create from file (title = filename) |
+| `ddctl update <ddocId> -f <filepath>` | Update from file |
+| `ddctl download <ddocId>` | Download document (`-o`/`--output`) |
+| `ddctl delete <ddocId> [...]` | Delete one or more documents |
+| `ddctl events list-failed` | List failed sync events |
+| `ddctl events retry <eventId>` | Retry one failed event |
+| `ddctl events retry-all` | Retry all failed events |
 
-Required body fields: onchainFileId, folderId, folderRef, folderName, portalAddress, metadataIPFSHash, lastTransactionBlockNumber, lastTransactionBlockTimestamp.
-Typically used for sync/on-chain integration, not ad-hoc folder creation by agents.
+---
 
-### Get Folder
-GET /api/folders/{folderRef}/{folderId}?apiKey={API_KEY}
-Both folderRef and folderId are required.
+# Response Schemas
 
-### Search Documents
-GET /api/search?apiKey={API_KEY}&q={query}&limit=10&skip=0
-Query parameter "q" is required.
+## Document
 
-Response:
+```json
 {
-  "nodes": [...],
-  "total": 100,
-  "hasNext": true
+  "ddocId": "string",
+  "title": "string",
+  "content": "string",
+  "syncStatus": "pending | synced | failed",
+  "link": "string | null",
+  "localVersion": 0,
+  "onchainVersion": 0,
+  "isDeleted": 0,
+  "onChainFileId": null,
+  "portalAddress": "string",
+  "createdAt": "ISO timestamp",
+  "updatedAt": "ISO timestamp"
 }
-(Search returns "nodes", not "ddocs".)
+```
 
-### List Failed Events
-GET /api/events/failed?apiKey={API_KEY}
+## List Response
 
-Response: array of events (e.g. _id, fileId, portalAddress, type, timestamp, lastError).
+```json
+{ "ddocs": [...], "total": 100, "hasNext": true }
+```
 
-### Retry All Failed Events
-POST /api/events/retry-failed?apiKey={API_KEY}
+## Search Response
 
-Response: { "retried": number }
+```json
+{ "nodes": [...], "total": 100, "hasNext": true }
+```
 
-### Retry One Failed Event
-POST /api/events/{id}/retry?apiKey={API_KEY}
+Note: search returns `nodes`, not `ddocs`.
 
-Response: { "ok": true } or 404 if event not found or not in failed state.
+## Create/Update Response
 
-## OpenAPI Specification
+```json
+{ "message": "...", "data": { "ddocId": "...", ... } }
+```
 
-A machine-readable OpenAPI 3.1 spec is available at GET {SERVER_URL}/openapi.json. Use it for auto-generating tool definitions or importing into Swagger UI / Postman.
+Extract ddocId from `data.ddocId`.
 
 ---
 
-# Agent Workflow Recipes
-
-## Recipe 1: Publish a local file to blockchain
+# Key Concepts
 
-1. Read the file content from the local filesystem
-2. POST /api/ddocs?apiKey={API_KEY} with JSON body: {"title": "<filename>", "content": "<file content>"}
-3. Extract ddocId from response.data.ddocId
-4. Poll GET /api/ddocs/{ddocId}?apiKey={API_KEY} every 5 seconds
-5. When syncStatus is "synced", show: "Published! [View Document](link)"
-6. If syncStatus is "failed", show: "Saved locally but blockchain sync failed. Run retry."
+- **DDoc**: A document with a unique `ddocId`
+- **syncStatus**: `pending` (saving to blockchain) -> `synced` (published, `link` available) or `failed`
+- **link**: Public URL to view the document on-chain (only present when synced)
+- **localVersion**: Increments on each local edit
+- **onchainVersion**: Version published to blockchain
 
-## Recipe 2: Find and update a document by title
+# Error Handling
 
-1. GET /api/search?apiKey={API_KEY}&q={title} to search for the document
-2. From the results (response.nodes), find the matching document and extract its ddocId
-3. PUT /api/ddocs/{ddocId}?apiKey={API_KEY} with JSON body: {"content": "<new content>"}
-4. Poll GET /api/ddocs/{ddocId}?apiKey={API_KEY} until synced
-5. Show the updated link
+| Status | Meaning |
+|--------|---------|
+| 400 | Validation error (missing/invalid params) |
+| 401 | Invalid or missing API key |
+| 404 | Resource not found |
+| 429 | Rate limited (respect Retry-After header) |
+| 500 | Server error (retry or inform user) |
 
-## Recipe 3: Check health and diagnose sync issues
+# OpenAPI Specification
 
-1. GET /ping - verify server is running (expect {"reply":"pong"})
-2. GET /api/ddocs?apiKey={API_KEY}&limit=5 - verify API key works and list recent docs
-3. GET /api/events/failed?apiKey={API_KEY} - check for failed sync events
-4. If failed events exist, POST /api/events/retry-failed?apiKey={API_KEY} to retry them
-5. Report: server status, number of documents, number of failed events (and if retried)
+A machine-readable OpenAPI 3.1 spec is available at `GET {SERVER_URL}/openapi.json`.
 
 ---
 
-# Claude Code Integration
-
-## Custom Slash Commands
+# Custom Slash Commands (Skills)
 
-You can create custom slash commands for Claude Code that integrate with Fileverse API. Create a `.claude/commands/` directory in your project and add markdown prompt files.
+If the user asks you to set up Fileverse slash commands, create the following files. These work in Claude Code and other tools that support markdown-based command prompts.
 
-### fileverse-publish.md
+Create a `.claude/commands/` directory in the project, then add these files:
 
-Create `.claude/commands/fileverse-publish.md`:
+## `.claude/commands/fileverse-publish.md`
 
 ```markdown
-Publish a file to Fileverse API blockchain storage.
+Publish a file to Fileverse blockchain storage.
 
 Steps:
-1. Read credentials from ~/.fileverseapirc (JSON with apiKey and serverUrl fields)
-2. Read the file at: $ARGUMENTS
-3. POST to {serverUrl}/api/ddocs?apiKey={apiKey} with JSON body:
-   Content-Type: application/json
-   {"title": "<filename without extension>", "content": "<file contents>"}
-4. Extract ddocId from response.data.ddocId
-5. Poll GET {serverUrl}/api/ddocs/{ddocId}?apiKey={apiKey} every 5 seconds until syncStatus is "synced" or "failed"
-6. If synced: show "Published! Link: {link}"
-7. If failed: show "Saved locally but blockchain sync failed"
+1. Read the file at: $ARGUMENTS
+2. Use the fileverse_create_document MCP tool with the filename (without extension) as the title and the file contents as content.
+3. If MCP tools are not available, fall back to the REST API:
+   - Read credentials from ~/.fileverseapirc (JSON with apiKey and serverUrl fields)
+   - POST to {serverUrl}/api/ddocs?apiKey={apiKey} with JSON body: {"title": "<filename>", "content": "<file contents>"}
+   - Extract ddocId from response.data.ddocId
+   - Poll GET {serverUrl}/api/ddocs/{ddocId}?apiKey={apiKey} every 5s until syncStatus is "synced" or "failed"
+4. If synced: show "Published! Link: {link}"
+5. If failed: show "Saved locally but blockchain sync failed"
 ```
 
 Usage: `/fileverse-publish path/to/document.md`
 
-### fileverse-list.md
-
-Create `.claude/commands/fileverse-list.md`:
+## `.claude/commands/fileverse-list.md`
 
 ```markdown
-List all documents in Fileverse API.
+List all documents stored in Fileverse.
 
 Steps:
-1. Read credentials from ~/.fileverseapirc (JSON with apiKey and serverUrl fields)
-2. GET {serverUrl}/api/ddocs?apiKey={apiKey}&limit=50
+1. Use the fileverse_list_documents MCP tool with limit 50.
+2. If MCP tools are not available, fall back to the REST API:
+   - Read credentials from ~/.fileverseapirc
+   - GET {serverUrl}/api/ddocs?apiKey={apiKey}&limit=50
 3. Display results as a table with columns: ddocId, title, syncStatus, link, updatedAt
 4. Show total count and whether more documents exist (hasNext)
 ```
 
 Usage: `/fileverse-list`
 
-### fileverse-search.md
-
-Create `.claude/commands/fileverse-search.md`:
+## `.claude/commands/fileverse-search.md`
 
 ```markdown
-Search for documents in Fileverse API.
+Search for documents in Fileverse.
 
 Steps:
-1. Read credentials from ~/.fileverseapirc (JSON with apiKey and serverUrl fields)
-2. GET {serverUrl}/api/search?apiKey={apiKey}&q=$ARGUMENTS
+1. Use the fileverse_search_documents MCP tool with query: $ARGUMENTS
+2. If MCP tools are not available, fall back to the REST API:
+   - Read credentials from ~/.fileverseapirc
+   - GET {serverUrl}/api/search?apiKey={apiKey}&q=$ARGUMENTS
 3. Display matching documents with: ddocId, title, syncStatus, link
 4. Show total matches found
 ```
 
 Usage: `/fileverse-search meeting notes`
 
-### fileverse-status.md
-
-Create `.claude/commands/fileverse-status.md`:
+## `.claude/commands/fileverse-status.md`
 
 ```markdown
-Check the sync status of a Fileverse API document.
+Check the sync status of a Fileverse document.
 
 Steps:
-1. Read credentials from ~/.fileverseapirc (JSON with apiKey and serverUrl fields)
-2. GET {serverUrl}/api/ddocs/$ARGUMENTS?apiKey={apiKey}
+1. Use the fileverse_get_sync_status MCP tool with ddocId: $ARGUMENTS
+2. If MCP tools are not available, fall back to the REST API:
+   - Read credentials from ~/.fileverseapirc
+   - GET {serverUrl}/api/ddocs/$ARGUMENTS?apiKey={apiKey}
 3. Display:
    - Title
    - Sync Status (pending/synced/failed)
    - Link (if synced)
    - Local Version vs On-chain Version
    - Last Updated
-4. If status is "failed", suggest running: POST {serverUrl}/api/events/retry-failed?apiKey={apiKey}
+4. If status is "failed", offer to retry using fileverse_retry_failed_events or POST {serverUrl}/api/events/retry-failed?apiKey={apiKey}
 ```
 
 Usage: `/fileverse-status <ddocId>`
 
----
+## `.claude/commands/fileverse-delete.md`
 
-# Key Concepts
-
-- **DDoc**: A document with unique ddocId
-- **syncStatus**: pending → synced (success) or failed
-- **link**: Public URL to view document (only available when synced)
-- **localVersion**: Increments on each edit
-- **onchainVersion**: Version on blockchain
-
-# Error Handling
-
-**400 Bad Request**: Validation errors (missing/invalid body or params). Response body: { "error": "..." } or { "message": "..." }.
-**401 Unauthorized**: Invalid or missing API key
-**404 Not Found**: Resource doesn't exist
-**429 Too Many Requests**: If returned (e.g. from upstream), respect Retry-After header if present.
-**500 Server Error**: Internal error, retry or inform user
-
----
+```markdown
+Delete a document from Fileverse.
 
-# Future Improvements (Planned)
+Steps:
+1. If $ARGUMENTS looks like a ddocId, use it directly. Otherwise, search for it:
+   - Use fileverse_search_documents MCP tool (or GET /api/search) with $ARGUMENTS as the query
+   - Show matching documents and ask the user to confirm which one to delete
+2. Use the fileverse_delete_document MCP tool with the ddocId.
+3. If MCP tools are not available:
+   - Read credentials from ~/.fileverseapirc
+   - DELETE {serverUrl}/api/ddocs/{ddocId}?apiKey={apiKey}
+4. Confirm deletion to the user.
+```
 
-- **SSE for Sync Status**: Real-time sync status push via GET /api/ddocs/:ddocId/stream to eliminate polling
-- **Batch Operations**: POST /api/ddocs/bulk for creating/updating/deleting multiple documents at once
-- **Partial Responses**: ?fields=ddocId,syncStatus,link for requesting only specific fields
-- **TypeScript SDK**: Typed client library with built-in sync polling at @fileverse/api/client
-- **Webhook Callbacks**: Register a URL to be notified when sync completes instead of polling
+Usage: `/fileverse-delete <ddocId or search term>`