Brevo skills
Give your AI agent step-by-step instructions for working with Brevo, then let it act through the Brevo MCP server or the Brevo CLI.
Overview
A skill is a short set of instructions (a SKILL.md file) that teaches an AI agent how to do one job well. Brevo skills cover jobs such as writing and checking an email, creating a campaign draft, sending a test, reading an account’s brand kit, or driving the Brevo Developer CLI.
Skills contain text only. They never hold an API key and never call an API directly. The email skills act on your account through the Brevo MCP server, which signs you in with OAuth, so there is nothing to copy or paste. The brevo-cli skill instead drives the Brevo Developer CLI from your shell, using the CLI’s own login.
The skills are published at getbrevo/brevo-skills and work with Claude, ChatGPT, Codex, Gemini, Cursor and any agent that reads SKILL.md. Anyone can browse, copy or fork the repository.
You can get them in six ways:
- Plugin: Claude Code, Codex, Gemini CLI, Cursor and Grok read the repository directly. ChatGPT has a published Brevo plugin.
- Brevo CLI:
brevo skill installcopies skills to your agent and checks every file against a hash. npx skills add: works with any agent that readsSKILL.md.- ZIP upload: for claude.ai, ChatGPT and the Gemini app, which have no git.
- Brevo MCP server: agents that can’t install anything load the skills through
get_brevo_guide. - Public repository: copy a skill by hand.
Available skills
The email skills work together: email routes a request to the others, and nothing is sent to real recipients without your explicit yes.
The full list and each skill’s version are in the catalog.
Install
Claude Code
Codex
Gemini CLI
Cursor
Brevo CLI
Any other agent
ChatGPT
claude.ai, Gemini app
The plugin adds the skills and the Brevo MCP server together.
Also supported: Claude Cowork (add the marketplace getbrevo/brevo-skills under Plugins), Grok (install the plugin from getbrevo/brevo-skills), Antigravity CLI (agy plugin install with the Antigravity ZIP from a release), Gemini Enterprise (import the skill ZIP) and the OpenAI Responses API (upload the skill ZIP to /v1/skills).
Each platform has its own limits on what it accepts. The platform guide lists every install path, what each one includes and the known limits. Use the plugin or the Brevo CLI for a given agent, not both, or the agent sees each skill twice.
Connect the Brevo MCP server
The email skills need the Brevo MCP server to act on your account (brevo-cli uses the CLI instead). The plugins above add it for you. For anything else, connect the endpoint for your platform and sign in with OAuth when your agent asks:
Every endpoint uses OAuth 2.1 with PKCE. Other MCP clients, such as VS Code and Windsurf, are covered in the MCP overview, which also lists the smaller per-module servers.
A skill is the playbook: what to check, what to ask you and in what order. The server’s tools do the work in your account. Skills on their own can’t reach your account, and the server on its own works but the agent has to work out the steps unaided. Tools can change between releases, so a skill names the tools it uses, and Brevo checks that they exist before each release.
Keep skills current
An agent that already has a skill uses it directly and doesn’t call the server for the same text. To check, an agent connected to the server calls list_brevo_guides with the versions it has. To update, tell your agent “update the Brevo skills”, or run the command for how you installed them:
No shell and no plugin
If your agent can only connect to the Brevo MCP server, it can still load the skills from the server. Two tools are built in:
list_brevo_guideslists the skills. Pass the versions you already have and each one comes back ascurrent,stale,ahead,unknownornot_installed, with a hint on how to update.get_brevo_guidereturns the full text of one skill. Your agent calls it only when the skill isn’t already loaded.
The brevo-guide skill ties these together. It tells the agent to use a skill that is already loaded, to load any other through get_brevo_guide, and where you can install the skills once if you want them without a tool call.
The server only reads and reports. It never installs or writes anything on your machine.
Download a skill without a shell
If your agent can fetch files but not run commands, it can install a skill from the release instead:
- Read the catalog from the latest release. Each skill that doesn’t need a shell has a
zip_url. - Download the skill’s ZIP from its
zip_url. - Check the file against
SHA256SUMS, published in the same release.
If you can’t fetch or check the file, use get_brevo_guide instead.
Use a skill
Ask for the job in plain language. Your agent picks the matching skill:
- “Read my Brevo brand kit and use it for my next email”
- “Create a Black Friday campaign draft for my VIP list”
- “Send a test of my welcome template to me@example.com”
- “Create a Brevo OAuth app with the CLI”
The skills create things as drafts where they can, and ask before anything you can’t undo, such as sending to real contacts.
About the skills repository
getbrevo/brevo-skills is where releases are published. Brevo writes and tests the skills in its own source repository, and each release copies them here:
skills/<name>/: the skills..claude-plugin/,.cursor-plugin/,.grok-plugin/,plugin.json,mcp.json,gemini-extension.json: the plugin files for each agent.catalog.json: the list of skills, versions and SHA-256 hashes that the Brevo CLI reads.- Each release also attaches the skill ZIPs and a
SHA256SUMSfile.
These files are overwritten at every release. The README, the guides in docs/ (platforms and how skills reach agents) and the ChatGPT plugin build files are edited in the repository by pull request. See the releases and the changelog for what changed.
Report a problem or contribute
- To suggest a change to a skill, open an issue on getbrevo/brevo-skills. The skills are written and tested in Brevo’s own source repository, so a pull request that edits a skill can’t be merged. Pull requests are welcome for the README and the docs in that repository.
- Report security problems privately, as described in the repository’s security policy.