DTC brands
Replace the photoshoot. Hero shots, lifestyle imagery, packshots, and product video on the same brief. No studio, no props, no day rate.
Loading…
Avocado AI · MCP server for any AI agent
Turn any AI agent into your creative engine
Connect Avocado AI to Claude, OpenAI Codex, Cursor, Hermes, OpenClaw or any MCP client, and generate product ads, images, video, music and voice from the conversation you already have open.
1
Open your agent's MCP settings
Claude: Settings, Connectors. Claude Code, Codex, Cursor, Hermes, OpenClaw: the MCP servers config.
Open connectors2
Add the Avocado endpoint
Paste the URL below. Claude signs you in; other clients take an API key as a Bearer token.
https://mcp.avocadoai.co/api/mcp3
Connect and sign in
Approve the connection, then try a prompt like: “Generate a 6-second Seedance video of a sneaker rotating on a black background.”
Everything you'd open four tabs for, in one chat.
Image, video, music, voice. Same conversation, same context, same brand voice. Claude keeps track of what you're making so iteration four still knows what you decided in iteration one.
Image generation
Pick a model, or let Claude pick one
Tell Claude what you're making and it routes to the right model. GPT Image 2 (currently #1 in the LMArena image leaderboard) when text on the can has to be readable. Nano Banana 2 for fast product iteration. Seedream 5 when you need clean photoreal at scale.
- 26+ image models, swappable mid-conversation
- 1:1 through 21:9, up to 4K
- Seeds preserved across the conversation. Regenerate frame 100 without losing frames 1 and 2
Video generation
Hero shots in the time it takes to brief them
Right now: Seedance 2.0, up to 1080p, available on every paid plan. We built it first because it's the model we'd pick ourselves for ad work. Fast turnaround, real cinematography, and it doesn't fall apart when you ask for camera moves.
- Seedance 2.0 and Seedance 2.0 Fast
- 5–15 seconds, up to 1080p
- More video models on the way. What you ship first should be the one you'd pick yourself
Music & SFX
The score and the SFX, on the same brief
Most ad teams skip music entirely or pay for stock libraries everyone else also uses. Generate it instead. Tell Claude the mood, the cut points, and the pace. Get back a track that's yours.
- Full tracks from 30 seconds to 5 minutes, generated to brief
- Targeted SFX for foley, hits, transitions, and ambience
- Commercial use, no watermarks, no licensing forms
Voice & narration
The voiceover, in 9 voices, in 30 seconds
Paste a script, pick a voice, get the take. The voices sound like 2026, not 2022 text-to-speech. Pair it with the video tool and Claude can hand you a UGC-style talking head from a written script alone.
- 9 voice presets, more on request
- Multiple languages
- Script honored exactly. No rewrites, no improvisation
Edit and remix
Edit by describing the change
Drop in a reference image, tell Claude what to change. “Swap the background to a beach.” “Put the product on a marble countertop.” “Make it look like the same brand shot in winter.” It runs the edit and hands the result back in the same chat.
- Nano Banana 2 for fast iteration, GPT-Image 2 when text matters
- Multi-image stitching, blending, reference-driven composition
- Each edit keeps the previous one in context. Layer changes without starting over
Built for the people already shipping ads
Most Avocado users are agencies, in-house creative teams, and DTC brands. People who get paid to ship, not to play.
Agencies & creative teams
Test 5 directions for the cost of 1. Pitch with finished comps, not stock-image mood boards. Win the brief, then keep the same chat open through delivery.
Freelancers & creators
Ship the post. Hook image, vertical video, voiceover, all in the time it used to take to write the caption.
Three things your agent can do here that single-tool generators can't.
Multi-asset briefs, iteration that builds on context, finishing in the same chat where you started. These are the moats.
1
Brief once
Brief once, generate everywhere
One prompt. Four assets. Same brand consistency across all of them because it's one conversation.
Example prompt
Five-second video of a sneaker rotating on a black background, plus three matching product stills, plus a 30-second voiceover script.
2
Iterate
Iterate on what you've already approved
Avocado remembers what you've made. Each iteration builds on the last. No re-uploading reference images, no re-explaining the brand.
Example prompt
Take the hero shot we just made and generate three alternates with the same lighting but different angles.
3
Finish
Multi-asset cuts, in one session
Claude pulls the asset URL, attaches it, and writes copy that matches the visual you just made. No copy-pasting, no asset wrangling.
Example prompt
Use that final video as the cover for the campaign, then write three caption variations.
Reference
What is the Avocado AI MCP server?
The Avocado AI MCP server is a Model Context Protocol server that lets any AI agent, including Claude, OpenAI Codex, Cursor, Hermes and OpenClaw, generate marketing images, videos, music and voice, edit assets, and publish posts through the Avocado AI creative workspace.
Endpoint: https://mcp.avocadoai.co/api/mcp. Transport: Streamable HTTP. Auth: OAuth for Claude connectors, an API key from your API keys page for every other client. Updated 2026-09-13.
Which AI agents work with it?
Any client that speaks Streamable HTTP MCP works, with no Avocado-specific plugin. The table lists the setup path for the agents we test against; a self-built agent follows the last row.
| Agent | Setup | Auth |
|---|---|---|
| Claude (web and desktop) | Settings, Connectors, add custom connector, paste the endpoint, sign in with your Avocado account | OAuth sign-in |
| Claude Code | claude mcp add avocado-ai --transport http <endpoint> with an Authorization: Bearer header | API key |
| OpenAI Codex | Add the endpoint as an MCP server in the Codex config with a Bearer header | API key |
| Cursor and Windsurf | Add the endpoint under MCP servers in settings with a Bearer header | API key |
| Hermes | Register the endpoint in the agent's MCP servers list; Avocado's own Super Agent runs on this path | API key |
| OpenClaw and any other MCP client | Point the client at the endpoint over Streamable HTTP and pass the API key as a Bearer token | API key |
How do I generate marketing images and videos from an AI agent?
Connect once, then brief the agent the way you would brief a designer. The agent chooses the tool, Avocado chooses the model, and the output is saved to your workspace with full commercial rights.
- Step 1
Connect
Add the endpoint to your agent (table above). Claude signs in; other clients use an API key.
- Step 2
Brief
"Generate a 9:16 TikTok ad of our protein powder on a gym bench, morning light." Or hand it a product photo and ask for a Seedance video.
- Step 3
Ship
The asset lands in your Avocado workspace. Ask the agent to edit it, upscale it, add it to a Flow, or publish it with the Connectors add-on.
What tools does the MCP server expose?
The server exposes more than 40 tools grouped into generation, brand, campaign canvas, publishing and account. Agents discover them automatically through the MCP tool list.
| Group | Tools |
|---|---|
| Generate | generate_image, edit_image, generate_video, generate_music, generate_sfx, generate_speech |
| Brand | extract_brand, apply_brand_layer, extract_product, remember, update_profile |
| Campaign canvas | create_flow, edit_flow, generate_image_to_flow, generate_video_to_flow, create_storyboard |
| Publish | publish_post, schedule_post, content calendar, comment automations, Meta Ads tools (with the Connectors and Operator add-ons) |
| Account | account_check_credits, models_list (free, no key needed), check_job |
Got any questions?
Common questions about connecting and creating through the Avocado MCP.
One URL away from a different workflow.
https://mcp.avocadoai.co/api/mcpPaste it into Claude → Settings → Connectors. Sign in with your Avocado account. Done.
New to Avocado? Start here.