About mtdt — Salesforce DevOps
mtdt is a remote MCP server for using Salesforce DevOps in MCP-capable coding agents. It supports metadata deployments, org comparisons and backups, with pre-deployment checks and role-based approval controls.
Key features
- 01Deploy metadata between Salesforce orgs
- 02Browse orgs and deployment sources
- 03Run pre-deployment checks and tests
- 04Compare orgs and back up metadata
- 05Move Work Items across environments
Use cases
- 01Promote metadata between Salesforce orgs
- 02Validate changes before deployment
- 03Compare org metadata or create backups
MCP support
Provides Salesforce DevOps tools to MCP clients through a remote Streamable HTTP endpoint with OAuth 2.0. Clients can browse metadata, validate and deploy changes, and access prompts for deployment and troubleshooting.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.mtdt.ioclaude mcp add --transport http io-mtdt-salesforce-devops https://mcp.mtdt.io{"mcpServers":{"io-mtdt-salesforce-devops":{"type":"http","url":"https://mcp.mtdt.io"}}}Open in Cursor{"servers":{"io-mtdt-salesforce-devops":{"type":"http","url":"https://mcp.mtdt.io"}}}Open in VS Code{"mcpServers":{"io-mtdt-salesforce-devops":{"type":"http","url":"https://mcp.mtdt.io"}}}{"mcpServers":{"io-mtdt-salesforce-devops":{"serverUrl":"https://mcp.mtdt.io"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEmtdt — MCP server for your AI agent
Salesforce DevOps in any MCP-capable coding agent. One remote endpoint:
https://mcp.mtdt.io
Streamable HTTP + OAuth 2.0 (browser sign-in, Dynamic Client Registration) — no API keys, no tokens to copy. Works with Claude Code, Cursor, OpenAI Codex, OpenCode, and anything else that speaks MCP.
What your agent gets
Deploy metadata between Salesforce orgs through mtdt.io: find orgs and
deployment sources (git repos, metadata backups, uploaded archives), create a deployment and edit
its components and test level, browse metadata, run the pre-deploy checks (missing dependencies,
target conflicts, profile & permission warnings, record-backup recommendations, Apex test
recommendations), validate with tests, and deploy: use execute_deployment for the current saved draft or execute_quick_deployment
for an explicitly selected validation package. A zip the agent already has (SFDX project or Metadata API package) can be uploaded straight to
mtdt and deployed whole. Compare two orgs, back up org metadata, and move Work Items through a project's
environments. Every launch answers at once with the run's state; your agent follows it with
get_deployment_run.
Deploy-capable tools require the Run (mtdt:deploy) consent scope and always respect your
team's mtdt roles — approval-gated targets stay approval-gated.
Claude Code plugin — connection + built-in skills
The plugin is more than the connection: it ships the deployment know-how as skills, so Claude doesn't just have the tools — it knows the playbook.
mtdt-deploy— the full promotion workflow: resolve the target safely (production targets get confirmed, never assumed), browse & pick components, the pre-flight battery (the same checks a human gets in the Deploy modal), validation-first flow, quick-deploy, deploying a local zip, and the guardrails: the 75% coverage rule, rollback snapshots, approval gates.mtdt-troubleshoot— reading failed runs: component vs test vs coverage failures, quick-deploy prerequisites and expiry, stuck retrievals, auth pitfalls — each with the fix, not just the diagnosis.
/plugin marketplace add mtdt-io/mcp
/plugin install mtdt@mtdt
On first use Claude Code opens your browser to sign in to mtdt (OAuth) — no config files.
Connect (any MCP client)
Verified copy-paste configs per client (and the type-field gotchas between them):
CONNECT.md. Every client receives the deployment playbook at connection time
(MCP instructions) and can pull the mtdt-deploy / mtdt-troubleshoot prompts; the plugin
above ships the same know-how as skills, applied without being asked.
Quick versions:
Notes
- Requires an mtdt.io account with at least one connected Salesforce org.
Privacy and support
The connector processes requested Salesforce metadata and deployment/backup information through the MTDT service. Scope consent, team roles and target approval policies apply to the signed-in account. Review the privacy policy and terms of service before connecting. For support, email [email protected] or use the contact form.
Directory packages
Maintainers can build the OpenAI/Codex and Claude distribution ZIPs with
python3 publishing/package.py build. Submission materials and instructions are in
publishing/README.md.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.