Your AI workflows,
connected to everything
Promptmark connects to external MCP servers and makes their tools available in your conversations and playbook steps. Browse tools, test them live, and call them from AI workflows — without writing integration code.
MCP server, meet MCP client.
Connect to any MCP server
Promptmark speaks the Model Context Protocol as a client. Point it at any MCP-compatible server — databases, SaaS platforms, internal tools — and it discovers every available tool automatically.
Use tools in conversations
Ask your AI to call a connected tool mid-conversation. Search a knowledge base, query a database, post to Slack — without leaving the chat. Tool results stream back in real time.
Wire tools into playbooks
Playbook steps can call tools from any connected server. Build multi-step workflows that reach across services — research, transform, deliver — all orchestrated through MCP.
Four steps. Zero glue code.
Add a server
Paste an MCP server URL and choose your auth method — bearer token, OAuth, or none. Promptmark encrypts credentials with AES-256 and stores them per-user.
Discover tools
Promptmark fetches the server’s tool list automatically. Browse tools with descriptions, input schemas, and connection status — all visible in the tool browser.
Attach to a conversation
Pick tools from any connected server and attach them to a conversation or test session. Your AI model sees them alongside Promptmark’s native tools.
Run in a playbook
Reference connected tools in playbook steps. When the step executes, Promptmark calls the external server, streams the result, and passes it to the next step.
Civic: 85 MCP servers through one connection
Civic Technologies is Promptmark's identity layer and its first featured MCP hub. Sign in with Civic Auth, then connect to Civic's MCP gateway — and 85 servers come online at once. Google Workspace, Slack, GitHub, Stripe, Firecrawl, and dozens more. One token. One connection. Automatic tool discovery.
Search emails, read spreadsheets, create tickets, post messages — tools for the services you already use.
Civic enforces access control and authorization policies on every tool invocation your AI makes.
New servers and tools appear as Civic adds them. No configuration changes on your end.
Tools that need user authorization trigger an in-app auth flow. Civic manages tokens and secrets — your agent never touches raw credentials.
Talk to external tools
Connected tools work like any other tool in a conversation. Ask your AI to use them and results stream back in real time.
"Search my Google Drive for the Q1 marketing brief and summarize the key metrics."
"Create a GitHub issue in the promptmark repo titled 'Update SSRF blocklist' with the label 'security'."
"Query the staging database for all users who signed up this week and format the results as a markdown table."
"Post a message to #engineering in Slack with the playbook execution summary from this morning's run."
"Use the Firecrawl tool to scrape the competitor's pricing page and extract their plan names and prices."
Connected, not exposed
Private IPs blocked
Connections to localhost, 10.x.x.x, 172.16-31.x.x, 192.168.x.x, and link-local ranges are rejected at validation time. No exceptions.
Cloud metadata blocked
AWS metadata endpoints (169.254.169.254), GCP, Azure, and DigitalOcean equivalents are all on the blocklist. Your cloud credentials stay invisible.
DNS rebinding blocked
Hostnames are resolved before connection, and the resolved IP is checked against the blocklist. A domain that resolves to a private IP is treated the same as a private IP.
Credentials are encrypted with AES-256-GCM per user. Connection pooling manages sessions automatically — reconnects happen without your intervention.
Bring your tools into the conversation
Free during beta. Connect any MCP server. Use tools in conversations and playbooks.
Create your free account