manus

# Manus AI Agent Delegate complex tasks to Manus - an autonomous AI agent that browses the web, writes code, generates reports, scrapes data, builds mobile apps, creates financial models, and delivers complete work products. ## API Configuration **Base URL:** `https://api.manus.im/v1` **Authentication Header:** `API_KEY: <your-key>` Set via: - `MANUS_API_KEY` environment variable - Or `skills.manus.apiKey` in OpenClaw config ## Agent Profiles and Pricing (Manus 1.6) > **IMPORTANT:** The `manus-1.5`, `manus-1.5-lite`, and `manus-1.5-max` profiles are DEPRECATED and will stop working. Always use the 1.6 profiles below. | Profile | Speed | Depth | Best for | Relative cost | |---------|-------|-------|----------|---------------| | `manus-1.6` | Standard | Full | Most tasks, file creation, reports | Standard (recommended default) | | `manus-1.6-lite` | Fast | Light | Quick lookups, simple questions, summaries | ~40% cheaper | | `manus-1.6-max` | Slow | Deep | Complex multi-step research, thorough analysis (+19.2% satisfaction vs standard) | ~2x standard | **Default:** Always use `manus-1.6` unless the user explicitly asks for lite or max. **Wide Research:** All Wide Research sub-agents automatically use `manus-1.6-max` for maximum depth. ### Cost estimation guidance Before creating a task, inform the user of the expected profile: - **Simple/quick tasks** (summaries, single-page lookups): recommend `manus-1.6-lite` - **Standard tasks** (reports, presentations, code projects): use `manus-1.6` - **Complex tasks** (multi-source research, competitive analysis, deep dives): suggest `manus-1.6-max` Users can override with "use lite", "use max", or "use the cheap one" / "use the thorough one". ## Manus 1.6 Capabilities Manus 1.6 adds these capabilities over 1.5: - **Mobile development** - build and preview mobile apps - **Design View** - visual design and layout capabilities - **Financial modeling** - spreadsheets, projections, financial analysis - **Telegram integration** - interact via Telegram bot - **Built-in connectors** - Gmail, Notion, Google Calendar, Slack, Similarweb ## Task Modes | Mode | Description | When to use | |------|-------------|-------------| | `agent` | Full autonomous agent with tool use | File creation, web browsing, code writing (recommended default) | | `chat` | Conversational back-and-forth | Simple Q&A, brainstorming | | `adaptive` | Auto-selects best approach | When unsure which mode fits | **Default:** Always use `agent` mode for tasks that produce deliverables. --- ## Task Templates Use these templates to create well-structured tasks. Always include `createShareableLink: true`. ### Research Template For market research, competitive analysis, topic deep-dives. ```bash curl -s -X POST "https://api.manus.im/v1/tasks" \ -H "API_KEY: $MANUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "Research [TOPIC]. Find the top [N] [entities] by [criteria]. For each, provide: name, description, key metrics, recent news. Compile into a structured report with sections, tables, and citations. Include a summary with key takeaways.", "agentProfile": "manus-1.6", "taskMode": "agent", "createShareableLink": true }' ``` **Example prompts:** - "Research the top 10 AI startups in healthcare by funding raised in 2025. For each provide founding year, total funding, key product, and notable customers." - "Do a competitive analysis of project management tools: Asana, Monday, Linear, Jira, ClickUp. Compare pricing, features, integrations, and user reviews." - "Research the current state of autonomous vehicle regulations in the US, EU, and China. Summarize key differences and recent changes." ### Report Generation Template For creating formatted documents, presentations, executive summaries. ```bash curl -s -X POST "https://api.manus.im/v1/tasks" \ -H "API_KEY: $MANUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "Create a [FORMAT] about [TOPIC]. Include: [SECTIONS]. Use professional formatting with headers, bullet points, and data tables where appropriate. Export as [FILE_TYPE].", "agentProfile": "manus-1.6", "taskMode": "agent", "createShareableLink": true }' ``` **Example prompts:** - "Create a 10-slide presentation about our company's AI strategy for the board meeting. Include market overview, our position, roadmap, and investment ask." - "Write an executive summary report on the state of remote work in 2025. Include statistics, trends, and recommendations. Export as PDF." - "Generate a one-page brief comparing AWS, GCP, and Azure for a startup choosing a cloud provider." ### Code Building Template For generating code projects, scripts, prototypes. ```bash curl -s -X POST "https://api.manus.im/v1/tasks" \ -H "API_KEY: $MANUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "Build a [LANGUAGE/FRAMEWORK] [PROJECT_TYPE] that [FUNCTIONALITY]. Include: [REQUIREMENTS]. Follow best practices for [LANGUAGE]. Include a README with setup instructions.", "agentProfile": "manus-1.6", "taskMode": "agent", "createShareableLink": true }' ``` **Example prompts:** - "Build a Python Flask API that takes a URL and returns a summary of the page content using BeautifulSoup for scraping. Include error handling, rate limiting, and a Dockerfile." - "Create a React dashboard component that displays real-time metrics from a WebSocket connection. Include TypeScript types and unit tests." - "Write a bash script that audits an AWS account for security misconfigurations: open S3 buckets, overly permissive IAM policies, unencrypted volumes." ### Data Scraping Template For extracting structured data from websites. ```bash curl -s -X POST "https://api.manus.im/v1/tasks" \ -H "API_KEY: $MANUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "Scrape [WEBSITE/SOURCE] and extract [DATA_FIELDS] for [ENTITIES]. Output as [FORMAT: CSV/JSON/table]. Include [N] results. Handle pagination if needed.", "agentProfile": "manus-1.6", "taskMode": "agent", "createShareableLink": true }' ``` **Example prompts:** - "Scrape the top 50 products from Product Hunt this week. Extract: name, tagline, upvote count, maker name, and URL. Output as CSV." - "Go to news.ycombinator.com and collect the top 30 stories. Extract title, URL, points, and comment count. Output as JSON." - "Find the top 20 open source AI projects on GitHub by stars in the last month. Extract repo name, description, stars, language, and last commit date." --- ## Core API Operations ### Create a Task ```bash curl -s -X POST "https://api.manus.im/v1/tasks" \ -H "API_KEY: $MANUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "Your task description here", "agentProfile": "manus-1.6", "taskMode": "agent", "createShareableLink": true }' ``` **Full parameter reference:** | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | `prompt` | string | (required) | The task description | | `agentProfile` | string | `manus-1.6` | Agent profile: `manus-1.6`, `manus-1.6-lite`, `manus-1.6-max` | | `taskMode` | string | `agent` | Task mode: `agent`, `chat`, `adaptive` | | `createShareableLink` | boolean | `false` | Generate a public share URL | | `interactiveMode` | boolean | `false` | Allow Manus to ask follow-up questions (task pauses with `stop_reason: ask`) | | `projectId` | string | - | Associate with a project's persistent instructions | | `taskId` | string | - | Continue an existing conversation (multi-turn) | | `locale` | string | - | Language locale (e.g., `"zh-CN"`, `"ja"`, `"es"`) | | `attachments` | array | - | Files to attach (see Attachment types below) | **Attachment types:** ```json // From an uploaded file ID (see Files API) { "type": "fromFileId", "fileId": "file_abc123" } // From a URL (Manus downloads it) { "type": "fromUrl", "url": "https://example.com/data.csv" } // Inline base64 content { "type": "fromBase64", "data": "base64-encoded-content", "fileName": "image.png", "mimeType": "image/png" } ``` **Response:** ```json { "task_id": "abc123", "task_title": "Task Title", "task_url": "https://manus.im/app/abc123", "status": "pending" } ``` After creating a task, immediately tell the user: 1. The task has been submitted 2. The task ID for reference 3. Expected duration (2-10+ minutes depending on complexity) 4. That you will monitor it for completion ### Multi-Turn Conversations To continue a previous task as a conversation, pass the original `taskId`: ```bash curl -s -X POST "https://api.manus.im/v1/tasks" \ -H "API_KEY: $MANUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "Now add a section about pricing trends", "agentProfile": "manus-1.6", "taskMode": "agent", "taskId": "PREVIOUS_TASK_ID", "createShareableLink": true }' ``` This preserves all context from the previous task. Use this when: - The user says "also...", "now add...", "refine the previous..." - A task completed with `interactiveMode` and Manus asked a question (`stop_reason: ask`) - The user wants to iterate on Manus output ### Get Task Details ```bash curl -s "https://api.manus.im/v1/tasks/{task_id}" \ -H "API_KEY: $MANUS_API_KEY" ``` **Status values:** | Status | Meaning | Action | |--------|---------|--------| | `pending` | Queued, not yet started | Wait and poll again in 15-30 seconds | | `running` | Actively working | Wait and poll again in 15-30 seconds | | `completed` | Finished successfully | Retrieve and deliver results | | `failed` | Task errored out | Report error, suggest retry | ### Update Task Metadata ```bash curl -s -X PUT "https://api.manus.im/v1/tasks/{task_id}" \ -H "API_KEY: $MANUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "title": "New title for this task" }' ``` ### Delete a Task ```bash curl -s -X DELETE "https://api.manus.im/v1/tasks/{task_id}" \ -H "API_KEY: $MANUS_API_KEY" ``` ### List Tasks (Paginated) ```bash curl -s "https://api.manus.im/v1/tasks?status=completed&page=1&limit=20" \ -H "API_KEY: $MANUS_API_KEY" ``` **Query parameters:** | Parameter | Type | Default | Description | |-----------|------|---------|-------------| | `status` | string | - | Filter by status: `pending`, `running`, `completed`, `failed` | | `page` | integer | `1` | Page number | | `limit` | integer | `20` | Results per page | When showing task history to the user, format it as a table: ``` | Task ID | Title | Status | Profile | Created |