autowhisper-mcp
This server provides an AutoWhisper AI CMO chat interface with confirmation handling for destructive actions.
Send natural-language instructions to the AutoWhisper CMO.
Target a specific product by
product_id.Add products via URL; the CMO extracts the product image, and text-only descriptions or placeholder/stock images are rejected.
Generate on-brand UGC videos, images, and social posts.
Approve content, connect social platforms, publish across 30+ networks, and report analytics.
Confirm or decline a destructive CMO action using its
message_idanddecision(yes/no).
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@autowhisper-mcp Add my product https://mystore.com/widget"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
AutoWhisper MCP
Turn your product into marketing content from the agent you already use. AutoWhisper gives your agent access to the same products, creative tasks, review queue, publishing connections and results as the web dashboard. Make UGC videos, images and social posts; review the result, then publish to the accounts you have connected.
Connect your agent to AutoWhisper · Explore the companion Skill
Start with one product
Connect this MCP server and authorize your AutoWhisper account.
Give your agent a product URL, or choose an existing product in a workspace.
Ask for a creative, inspect its preview and cost, then decide whether to render or publish.
Try these prompts:
“Show my products and connected accounts. Don’t start any generation.”
“For this product, prepare one UGC video. Show the preview and quoted cost before continuing.”
“Show which posts failed and how the published content performed.”
Related MCP server: Ayrshare MCP Server
Connect by URL
For clients that support remote HTTP MCP and its authentication, use:
https://autowhisper.xyz/mcpIn Claude Code:
claude mcp add --transport http autowhisper https://autowhisper.xyz/mcpThen open /mcp, select AutoWhisper and authenticate. Other compatible clients use their own connector setup. Authorize on the AutoWhisper page; you do not need to paste a token into a conversation. An AutoWhisper account is required.
Prefer a Claude Code plugin that installs the Skill and MCP together? Follow the companion Skill quick start.
Local stdio alternative
For clients that need a local process, Node.js 18+ is required:
{
"mcpServers": {
"autowhisper": {
"command": "npx",
"args": ["-y", "autowhisper-mcp"]
}
}
}Ask the agent to connect to AutoWhisper. autowhisper_connect returns an approval link and picks up authorization without a restart. Skill and local MCP share ~/.config/autowhisper/credentials.json.
Already have an API token? Set AUTOWHISPER_API_TOKEN in the server environment or secret store. Keep it out of chat and source control.
Setting | Default | Purpose |
| Optional | Existing token; otherwise use the approval flow. |
|
| Override the shared credential-file path. |
|
| Self-hosted or staging API origin. |
What the tools do
Task | Tools |
Connect and choose a workspace |
|
Read product facts, materials and presets |
|
Create, revise, resume or render an explicit content request |
|
Plan marketing or discuss an ambiguous creative request |
|
Review content and inspect delivery |
|
Edit text or take an explicit action |
|
Check credits and results |
|
Use the read tools for facts; they avoid a paid CMO chat turn. Explicit supported content requests use context → create/update → confirmation/review → task query. Marketing judgment can use CMO consultation. All four direct-content tools require workspace_id; writes also require a stable request_id. Check the server-returned capabilities and next actions: a tool in source or a package does not guarantee it is enabled for every account.
Preview, rendering and publishing are separate
acceptedacknowledges a request;appliedmeans it reached the native task. Onlydelivery_complete=trueconfirms the requested deliverable is ready. A script or first frame is not a finished video.Generation and CMO consultation use AutoWhisper credits. Read the quote and actual
billing; charges, refunds and included usage are distinct. Inspect your wallet for your account’s current balance and terms.Approving a feed item can render, charge and publish to connected platforms. Follow the returned confirmation. With no compatible account connected, it schedules nothing; connecting later does not retroactively publish it.
On an uncertain response, query the original
request_idor CMOmessage_idbefore retrying. Reuse identical request payloads and IDs; do not start another charged turn to recover the first.Use the server-returned preview approval and current revisions when resuming or rendering.
previous_previewsis history; current partial artifacts are not a completed result.AutoWhisper can prepare content and marketing advice. Publishing does not guarantee traffic, sales or ad-campaign execution. Competitor research occurred only if the actual task returns evidence of it.
For the detailed workflow and API reference, see the AutoWhisper Skill.
MCP or Skill?
MCP supplies callable tools; the Skill teaches an agent when and how to use the workflow. Use MCP alone for tool access, or both for the guided product → preview → review → publish workflow. They use the same AutoWhisper account and native tasks.
Start with your product on AutoWhisper
Development
npm ci
AUTOWHISPER_CONTRACT_FILE=test/fixtures/agent_content_contract.json npm testThis builds TypeScript and runs offline contract/behavior checks. Regenerate the fixture from the Rails contract test (UPDATE_AGENT_CONTENT_CONTRACT=1) and src/content-contract.ts together when changing the API. A passing fixture is separate from production rollout or a real generation/publishing test.
License
MIT. The open-source license covers this client; hosted AutoWhisper usage follows the service’s account terms and credits.
Available Tools
2 toolsautowhisper_cmoTalk to your AutoWhisper CMOAInspect
Send a natural-language instruction to your AutoWhisper AI CMO and get its reply. The CMO can add a product, generate on-brand content (UGC video, posts, images), approve items, connect social platforms, publish across 30+ networks, and report analytics. Examples: "Add my product https://mystore.com/widget and start the CMO", "Generate a UGC video for my product", "Publish the approved content", "How did last week's posts do?". To add a product, pass a product URL (the CMO extracts the image from the page) — a text-only description will not create it, and placeholder/stock images are rejected.
Inputs: instruction (required), product_id.
autowhisper_confirmConfirm an AutoWhisper actionAInspect
Approve or decline a destructive action the CMO asked you to confirm (surfaced by autowhisper_cmo). Pass the message_id it gave you and decision "yes" or "no".
Inputs: decision (required), message_id (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 2 tools
Each tool has a distinct purpose: one sends instructions to the AI CMO and retrieves responses, the other confirms destructive actions. There is no overlap or ambiguity.
Both tools follow the same naming pattern: 'autowhisper_' prefix with a descriptive verb ('cmo', 'confirm'). The convention is consistent and predictable.
With only 2 tools, the set feels minimal for a marketing automation assistant. While the single 'cmo' tool is highly capable, the count is borderline and could benefit from more specialized tools.
The tool set lacks granular operations for managing products, content, or analytics independently. Users must rely on unstructured natural language instructions to the CMO, creating potential dead ends and reliance on the CMO's inference.
Maintenance
Related MCP Connectors
Post to social media, run Google Ads, manage Shopify, inventory and phone calls from your AI.
Turn product pages and creative briefs into finished video and image ads from your AI chat.
Create, manage, schedule, and publish short-form user-generated content through AI agents.
Build and run AI image, video and on-brand ad workflows from chat, with brand kits and share links.
Related MCP Servers
- AlicenseAqualityDmaintenanceAI-powered social media posting across 14 platforms. Post to Twitter, Instagram, TikTok, Facebook, LinkedIn, YouTube and more with one command. AI adapts content per platform, schedules posts, and generates 30-day content calendars.6MIT
- AlicenseNot gradedqualityDmaintenanceEnables posting and managing content across 13+ social media platforms with scheduling, analytics, AI generation, and approval workflows through natural language.2MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to automate social media content generation, approval, and publishing, along with blog management, SEO/GEO audits, and ad operations through natural language.3AGPL 3.0
- AlicenseNot gradedqualityCmaintenanceConnects AI agents to Shopify, Meta Ads, Google Ads, and TikTok Ads for cross-platform reporting, creative management, and approval-gated campaign and store changes.MIT