Connect AgentReel to your AI tools (MCP)
AgentReel runs a remote MCP (Model Context Protocol) server at https://agentreel.co/api/mcp. Add that one address to Claude Code, Claude, ChatGPT, Codex or Cursor, sign in with your AgentReel account, and your AI app can make images, AI videos, template videos, video ads, voiceovers, music, Shorts and B-roll cuts for you. It uses the same tools, models and credits as the website. This guide shows the setup for each app, the tools your app gets, how credits and limits work, and how to disconnect an app. The MCP server page has the short overview.
Is it free? Connecting is free, and you never paste an API key. Before your first purchase, a connected app can spend credits on Z-Image images, video templates and Code to Video only. Every other tool needs a purchase. The free B-roll trial and the free AI video are only on the website.
Quick answers
What is the AgentReel MCP server?
It is a remote server that speaks MCP, the open standard AI apps use to call tools on other services. Your AI app connects to it over the internet and gets 42 tools: 31 AgentReel creative tools, 10 helpers for files, jobs, your balance, your B-roll projects, the editing skill library and this help center, and one tool that runs a skill of that library.
Which apps can connect?
Any app that supports remote MCP servers over Streamable HTTP with OAuth sign-in. This guide covers Claude Code, Claude (claude.ai and Claude Desktop), ChatGPT, Codex and Cursor.
Do I need an API key?
No. The app opens AgentReel in your browser, you sign in, and you choose Allow or Deny. AgentReel asks every time an app connects, and the app never sees your password.
What does it cost?
Connecting costs nothing. Each tool spends the same credits as on its own page, from the same balance, and a failed job gives its credits back. The section Credits and limits below has the details.
Where do my files go?
Each connected app gets its own workspace: a conversation named after the app in the Agent. Every file the app imports and every finished result is saved there.
How do I disconnect an app?
Open Connected apps and select Disconnect next to the app. It loses access at once.
Before you start
- A registered account. Sign in with Google, GitHub or an email code. A guest session cannot connect an app: register first, then connect.
- The address. Every app uses the same URL:
https://agentreel.co/api/mcp. - A browser on the same computer. The first time the app calls AgentReel, your browser opens the AgentReel sign-in and consent page. Select Allow, and the browser sends you back to the app.
Claude Code
- In a terminal, run
claude mcp add --transport http agentreel https://agentreel.co/api/mcp. Add--scope userbefore the name to use the server in every project, not only the current one. - Start Claude Code, type
/mcp, select agentreel and choose to authenticate. - Your browser opens AgentReel. Sign in, check the app name, and select Allow.
- Back in Claude Code, ask for what you want, for example "make a 9:16 product video from these three photos with AgentReel".
Claude Code keeps the sign-in and renews it by itself. Run /mcp again to see the status or to sign in again.
Claude and Claude Desktop
The connector you add on claude.ai also works in Claude Desktop and the Claude mobile apps, because they share your Claude account.
- On claude.ai or in Claude Desktop, open Customize, then Connectors.
- Select Add custom connector, type a name such as AgentReel, paste
https://agentreel.co/api/mcpas the URL, and select Add. - Select Connect next to the new connector. Sign in to AgentReel in the window that opens and select Allow.
- In a chat, turn on the AgentReel connector from the tools menu of the message box, then ask for your image or video.
The Free plan of Claude allows one custom connector. On Team and Enterprise plans, an Owner adds the connector in Organization settings, then Connectors, and each member connects their own AgentReel account.
ChatGPT
ChatGPT connects to remote MCP servers that use OAuth sign-in. Where you add one, and whether you can, depends on your ChatGPT plan and, in a business workspace, on what your admin allows.
- On chatgpt.com, open the plugins page (
chatgpt.com/plugins), select +, then Add custom MCP server. In the ChatGPT desktop app, the same option is in Settings, then MCP servers. - Name it AgentReel, paste
https://agentreel.co/api/mcpas the server URL, and choose OAuth as the authentication. - ChatGPT opens AgentReel. Sign in and select Allow. AgentReel then sends you back to ChatGPT.
- In a chat, turn on AgentReel from the tools menu and ask for what you want.
If you do not see the option, your plan or workspace does not offer custom MCP servers yet. The other apps in this guide work the same way with the same account.
Codex
- In a terminal, run
codex mcp add agentreel --url https://agentreel.co/api/mcp. - Run
codex mcp login agentreel. Your browser opens AgentReel: sign in and select Allow. On a computer without a browser, add--no-browserand open the printed link on another device. - Run
codex mcp listto check that agentreel shows as signed in. - In Codex, ask for an image or a video, and Codex calls the AgentReel tools.
Cursor
- Open
~/.cursor/mcp.json(create it if it does not exist) and add the server:{ "mcpServers": { "agentreel": { "url": "https://agentreel.co/api/mcp" } } }. If the file already lists other servers, add only theagentreelentry insidemcpServers. - Open the Cursor settings, find agentreel in the MCP section, and turn it on.
- Select the sign-in button when Cursor shows it. Sign in to AgentReel in your browser and select Allow.
- In the Cursor chat, ask the agent for what you want, for example a short demo video of your app.
To use AgentReel in one project only, put the same file at .cursor/mcp.json inside that project.
Tools
The server gives your app 42 tools. The first 31 make or find something and start a job; the next 10 are helpers that charge nothing; the last one starts a job with a skill of the editing skill library.
| Tool | What it does | Before your first purchase |
|---|---|---|
generate_image | An image from a prompt and optional reference images, with any model of the AI Image Generator | Z-Image only |
generate_video | A video clip from a prompt or a start image, with any model of the AI Video Generator | Needs a purchase |
template_video | A render of one of the AI video templates from its fields | Yes, with free credits |
lyric_video | A lyric video over your own song, with the lyrics timed word by word, as on the Lyric Video Maker. You confirm that you may use the song | Needs a purchase |
template_audio_video | A video template rendered with an AI voiceover, a music bed or both, and captions timed to the voice | Needs a purchase |
code_video | An original motion graphics video written in code, as on Code to Video | Yes, with free credits |
html_render | An MP4 render of an HTML animation you already wrote | Needs a purchase |
tiktok_carousel | A TikTok photo carousel of 3 to 8 slides from a topic, with its caption and hashtags, as on TikTok Carousel Maker. It does not post it | Needs a purchase |
translate_video | Subtitles or a dubbed voice in another language, as on the Video Translator | Needs a purchase |
avatar_video | A portrait that speaks your script, as on AI Avatar Video | Needs a purchase |
clone_voice | A clone of your own voice from a short recording | Needs a purchase |
text_to_speech | A voiceover of your text as an audio file, in a voice you choose | Needs a purchase |
generate_music | A new piece of music from a text prompt, as an MP3 of 3 to 300 seconds | Needs a purchase |
sound_effect | One sound effect from a short description, as an MP3 of up to 30 seconds | Needs a purchase |
isolate_voice | The clean voice of a recording of up to 30 minutes, without its music and background noise | Needs a purchase |
transcribe_speech | A transcript of a recording of up to 2 hours that tells the speakers apart, with an SRT subtitle file | Needs a purchase |
text_to_dialogue | A conversation of up to 40 lines in up to 10 voices as one MP3, with an SRT subtitle file of the lines | Needs a purchase |
change_voice | The speech of a recording of up to 5 minutes spoken again in another voice, with its timing and tone | Needs a purchase |
align_transcript | A script you already have timed word by word to a recording of up to 50 minutes, as a word list and an SRT subtitle file | Needs a purchase |
design_voice | A new voice from a description, kept among your own voices for later voiceovers | Needs a purchase |
find_voice | The top shared voices of the Voice Library for a style, language, gender and accent, with names and preview links; the best one is ready for a voiceover | Free |
split_stems | A song split into vocals and instrumental, or into six stems, as one MP3 each | Needs a purchase |
score_video | Music composed to the picture of a video of up to 10 minutes | Needs a purchase |
video_ad | A finished video ad from a product brief, as on the AI Video Ad Generator | Needs a purchase |
talking_head_cut | A jump cut of a talking-head recording, as on the Talking-Head Cutter | Needs a purchase |
make_shorts | Up to 5 vertical Shorts of 15 to 60 seconds from one recording of up to 90 minutes | Needs a purchase |
edit_video | A video edited by one instruction, as in the AI Video Editor | Needs a purchase |
video_remix | Why a reference video of up to 3 minutes holds attention and, for your topic, a new script and storyboard in its style, as on Video Remix | Needs a purchase |
find_highlights | Up to 5 moments of a video that match what you ask for, with their start and end times | Needs a purchase |
broll_cut | A captioned B-roll video from a script and your clips, as in the AI B-Roll Editor | Needs a purchase |
broll_revise | Changes a finished B-roll cut from your own words. Only the changed parts are rendered and charged | Needs a purchase |
import_file | Copies a file from a public link into the app's workspace | Free |
create_upload | Prepares an upload of a file from your computer into the workspace | Free |
complete_upload | Finishes that upload and adds the file to the workspace | Free |
list_files | Lists the files of the app's workspace | Free |
get_job_status | Reads a job and returns its files when it is done | Free |
list_jobs | Lists the app's recent jobs | Free |
get_credits_balance | Reads your credit balance and the free tier rule | Free |
search_help_docs | Searches this help center | Free |
list_library_skills | Lists the editing skills run_library_skill can run, such as captions, silence and scene cuts, reframing and mixing, and reads the inputs of one skill. The list changes as skills are added | Free |
list_broll_projects | Lists your B-roll projects, newest first, so broll_revise can change one | Free |
run_library_skill | Runs one skill of that list as a job, charged like the same skill in the Agent | Needs a purchase |
Each tool tells your app its inputs and its models, so you can ask in plain words. Your app picks the tool and fills in the details.
Files and jobs
- Your workspace. Each connected app works in its own conversation in the Agent, named after the app. Open it on the website to see, download or reuse every file.
- Bring a file. Give your app a public
httpslink, andimport_filecopies the file into the workspace. It takes MP4, MOV, PNG, JPG, WEBP, MP3, WAV, M4A and SRT files of any size, with no daily upload limit. A file on your computer comes in the same way throughcreate_uploadandcomplete_upload. You can also upload a file to the app's conversation on the website; the app then finds it withlist_files. - Jobs. A creative tool starts a job and answers with a
job_idand its estimated credits. Your app then callsget_job_statusuntil the job is done. Withwait_seconds(up to 25) one call waits for the result, so the app asks less often. - Results. A finished job returns the links of its files, which are also saved in the workspace.
Credits and limits
A connected app spends the credits of your AgentReel account, at the same prices as each tool's page. Your app can read your balance with get_credits_balance before it starts.
- Free accounts. Before your first purchase, a connected app can spend credits on Z-Image images, video templates and Code to Video only. Every other tool needs a purchase. The free B-roll trial and the free AI video are only on the website. The daily check-in gives 15 free credits once per UTC day on the website.
- Paid accounts. Every tool and every model is open. See the pricing page and Credits, plans and free usage.
- Refunds. A job that fails gives its credits back to the pool they came from.
- Limits per account. 60 calls a minute, 1,000 charged tool calls a day, and 5 running jobs at a time. When a limit is reached, the tool answers with an error your app can read, and you can try again later.
When a tool needs a purchase or more credits, it answers with PURCHASE_REQUIRED, FREE_MODEL_ONLY or INSUFFICIENT_CREDITS and a link to the pricing page, and nothing is charged.
Disconnect
- Open Connected apps in your dashboard to see every app you allowed, when it connected and when it was last used.
- Select Disconnect to end the access of an app at once. Removing the server inside the app alone does not end it.
- A sign-in stays valid for one hour, and the app renews it by itself as long as it is used at least once every 30 days. A connection ends after 180 days at the latest, and then the app asks you to sign in again.
Privacy
A connected app can do only what you allowed on the consent page:
- run AgentReel tools that spend your credits,
- import, list and read the files of its own workspace,
- read its jobs and your credit balance,
- list your B-roll projects and change a cut,
- search the help center.
It never sees your password or your payment details, and it cannot read your other Agent conversations. Of your projects, it can list only your B-roll projects. The prompts and files you send through the app are also handled by the app's own provider under its privacy terms. Read how AgentReel handles your data in Privacy and safety.