Audryo MCP server
Commit Nov 14, 2026Allows AI agents to manage lifecycle email operations including audiences, journeys, campaigns, and transactional sends through a safety-first sandbox.
Also known as Audryo/agent-toolkit, agent-toolkit
- Stars
- 2
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- 0.2.0
- Last commit
- Nov 14, 2026
- License
- MIT
What is Audryo MCP?
Audryo allows AI agents to manage lifecycle email operations including audiences, journeys, campaigns, and transactional sends through a safety-first sandbox.
Summary generated from the project's README
About Audryo MCP server
Audryo provides a specialized toolkit for AI agents to manage lifecycle email operations including audiences, journeys, campaigns, and transactional sends. Developers can connect this MCP server to clients like Claude Code, Codex, or Cursor using an API key and a project ID. The system operates through a safety-first model where agents create proposals and drafts by default, ensuring that no communication reaches a customer without human approval.
Users can configure specific permissions using different API key scopes. For instance, a read-only key allows for inspection, while a key with the deliver scope is required to activate journeys or send campaigns. The server includes a sandbox environment where agents can test real journeys end to end by capturing emails in a sandbox inbox. This setup allows for dry runs and fast-forwarding journeys to verify logic before going live. Every tool available in the MCP server is also accessible via a CLI for direct command-line interaction.
What you can do with it
- 1Automate onboarding journey setupAn agent can build an audience, draft a multi-email journey, and run a dry run in the sandbox for review.
- 2Test email workflows safelyDevelopers can use the sandbox to verify that new projects and journeys function correctly before deployment.
- 3Manage contact data and eventsAn AI client can ingest contact lists and event data to populate the email platform.
- 4Draft marketing campaignsAgents can propose new campaigns by selecting specific audiences and defining subject lines.
Written from the project's README. Every use case is checked against a quote from it.
How to install Audryo MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall Audryo MCP in Claude Code
$ claude mcp add --transport http agent-toolkit https://mcp.audryo.com/mcpInstall Audryo MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"agent-toolkit": {
"url": "https://mcp.audryo.com/mcp"
}
}
}Install Audryo MCP in VS Code
.vscode/mcp.json
{
"servers": {
"agent-toolkit": {
"type": "http",
"url": "https://mcp.audryo.com/mcp"
}
}
}Install Audryo MCP in VS Code CLI
$ code --add-mcp '{"name":"agent-toolkit","type":"http","url":"https://mcp.audryo.com/mcp"}'Install Audryo MCP in Codex
$ codex mcp add agent-toolkit --url https://mcp.audryo.com/mcpInstall Audryo MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.agent-toolkit] url = "https://mcp.audryo.com/mcp"
Install Audryo MCP in Gemini CLI
$ gemini mcp add -s user --transport http agent-toolkit https://mcp.audryo.com/mcpInstall Audryo MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"agent-toolkit": {
"url": "https://mcp.audryo.com/mcp",
"type": "http"
}
}
}Configuration
AUDRYO_API_KEYAUDRYO_PROJECT_ID
| Name | Description | |
|---|---|---|
| AUDRYO_API_KEY | — | |
| AUDRYO_PROJECT_ID | — |
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableaudryo/agent-toolkit is public and reachablegithub.com/audryo/agent-toolkit ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-10-24github.com/audryo/agent-toolkit ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeHTTPS endpoint answered 401 (sign-in required)Pass
README
View on GitHubAudryo agent toolkit
Let your AI agent run your lifecycle email: audiences, journeys, campaigns, emails and transactional sends in Audryo, through the same API the app uses. Your agent does the setup work; nothing reaches a customer until a person approves it.
This repo ships the audryo CLI, an MCP server and agent skills for
Claude Code, Codex, Cursor and any other MCP client.
You: Set up an onboarding journey for signups who haven't created a
project after 3 days.
Agent: Reads your data schema, builds the audience, drafts a three-email
journey, dry-runs it and fast-forwards it in the sandbox. Then it
hands back what to review. Nothing has been sent.
Safety model
- Drafts by default. Journeys, emails and campaigns an agent creates are review-only proposals.
- Sending needs
deliver. Without that scope a key cannot activate a journey or send anything, even by mistake. With it, Audryo still requires a passing dry run and confirmed emails. - Sandbox first. New projects capture every email in a sandbox inbox, so agents can test real journeys end to end. Going live is a human decision in the app.
- Retry-safe. Sends and proposals take an
Idempotency-Key, so a retried call never sends twice. - No inferred consent. Ingesting contacts never grants marketing consent.
Quick start
- Sign up (free during the public beta) and create a project. It starts in the sandbox.
- Create an API key under Settings → Developer with the smallest scope that works (see Configuration).
- Connect your agent:
Claude Code (plugin: MCP server + skills)
export AUDRYO_API_KEY="sf_live_..."
claude plugin marketplace add Audryo/agent-toolkit
claude plugin install audryo@audryo
Any agent: skills only
npx skills add Audryo/agent-toolkit
Any MCP client
{
"mcpServers": {
"audryo": {
"command": "npx",
"args": ["-y", "audryo", "mcp"],
"env": {
"AUDRYO_API_KEY": "sf_live_...",
"AUDRYO_PROJECT_ID": "<project-uuid>"
}
}
}
}
Codex: codex mcp add audryo --env AUDRYO_API_KEY=sf_live_... -- npx -y audryo mcp
Configuration
| Variable | Required | Default |
|---|---|---|
AUDRYO_API_KEY |
yes | — |
AUDRYO_PROJECT_ID |
for project tools | list with npx audryo bootstrap |
AUDRYO_API_BASE |
no | https://api.audryo.com/v1 |
Give agents the smallest key that works:
| Task | Scopes |
|---|---|
| Inspect | read |
| Sync data, draft journeys, emails and campaigns | read, write |
| Activate journeys, send campaigns and transactional email | read, write, deliver |
MCP tools
| Area | Tools | Scope |
|---|---|---|
| Orient | get_manifest, get_bootstrap, get_setup_context, get_brief, get_plan_usage |
read |
| Product and data | get_product_context, get_data_context, list_contacts, ingest_contacts, ingest_events |
read / write |
| Audiences | list_audiences, get_audience, preview_audience, create_audience, update_audience, delete_audience, suggest_audiences |
read / write |
| Journeys | list_journeys, inspect_journey, create_journey_proposal, revise_journey, propose_from_finding, run_dry_run |
read / write |
| Emails | create_email, create_journey_email |
write |
| Campaigns | list_campaigns, draft_campaign |
read / write |
| Sandbox | get_delivery_mode, list_sandbox_messages, get_sandbox_message, fast_forward_journey |
read / write |
| Insights | get_delivery_health, get_performance, get_observation, run_observation, get_copilot_plan |
read |
| Go live | activate_journey, pause_journey, send_campaign, send_transactional |
deliver |
Every tool is also a CLI command.
CLI
npx audryo bootstrap
npx audryo contacts ingest contacts.json
npx audryo events ingest events.json
npx audryo journeys list
npx audryo journeys inspect <workflowId>
npx audryo audiences create --name "All contacts" --definition '{"operator":"and","conditions":[]}'
npx audryo campaigns draft --name "July update" --audience-id <id> --subject "What's new"
npx audryo delivery health
Skills
| Skill | Teaches the agent to |
|---|---|
audryo-api |
Connect, pick scopes, use MCP or CLI instead of curl, stop at review |
audryo-email |
Build valid email documents, handle locales, save drafts for review |
audryo-sending |
Choose communication classes, respect consent, protect deliverability |
Full guide: https://app.audryo.com/docs/agents/quickstart
Contributing
Issues and pull requests are welcome. See CONTRIBUTING.md and the changelog. Report security issues privately: SECURITY.md.
License: MIT
Audryo MCP questions
How do I add Audryo MCP to Claude Code?
Run claude mcp add --transport http agent-toolkit https://mcp.audryo.com/mcp in your terminal. Add --scope user to make it available in every project.
How do I add Audryo MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
Can an AI agent send emails to customers automatically?
No, agents create review-only proposals by default. Sending requires a specific 'deliver' scope, a passing dry run, and human approval.
How can I test email journeys without sending them to real people?
You can use the sandbox mode where new projects capture every email in a sandbox inbox for end-to-end testing.
Does ingesting contacts automatically grant marketing consent?
No, the system does not infer consent when ingesting contact data.
How does the system prevent duplicate emails if a command is retried?
The system uses an Idempotency-Key for sends and proposals to ensure retried calls do not result in duplicate sends.