AgentKit by Scalekit provides an MCP gateway that manages user authentication, scopes tool access, and logs tool execution for AI agents.
What the product does and how it is positioned
AgentKit by Scalekit serves as a managed Model Context Protocol (MCP) gateway that connects AI agents to external applications and APIs. It manages credential storage, token refreshes, and tenant isolation so that agents execute actions as authenticated end users without exposing access tokens to agent prompts or application code.
The platform supports over 600 connectors and 30,000 tools, enabling developers to build Virtual MCP Servers that expose only the necessary tools to specific agents. It includes built-in execution triggers, custom connector extensions, attributed audit logging, and SIEM streaming for enterprise observability.
Source-supported ways to use the product
The official customer story reports that Von deployed agents that interact with Salesforce, Gong, and Google Drive on behalf of customers without rebuilding credential infrastructure.
The company presents customer implementations, such as Z47, using Scalekit's identity layer to establish discrete agent identities and control tool availability per role.
Developers use triggers to automatically start agent runs when upstream application events occur, such as a deal change or incoming email, operating directly under the connected user's permissions.
AgentKit operates as an intermediary gateway between AI agent frameworks and third-party SaaS services. By abstracting tool interaction behind an MCP endpoint, it performs permission checks and handles credential injection for each outbound request. User tokens are kept in an encrypted vault, shielding credentials from developer code, context windows, and model prompts.
For enterprise security environments, Scalekit offers self-hosted deployment options, including inside private VPCs or fully air-gapped infrastructures. Organizations can manage their own encryption keys via AWS KMS or GCP KMS, enforce US or EU data residency at the storage layer, and restrict operational access through IP allowlisting and real-time SIEM log streaming.
Checks to run with your own material and workflow
What was checked and when
Answers based on the source-checked product record
AgentKit by Scalekit is an MCP gateway that connects AI agents to external tools and APIs with managed authentication, scoping, and logging.
Users link their own accounts through supported auth models like OAuth 2.0, API keys, or bearer tokens, which the gateway stores encrypted and refreshes automatically.
Virtual MCP Servers are curated subsets of tools from connected applications, exposed via a single endpoint to limit context overhead and scope agent access.
Yes, tool execution logs contain user, agent, and tool attribution and can be exported or streamed directly to Datadog, Splunk, or any SIEM.
Yes, AgentKit supports private deployment inside a customer VPC, self-hosted setups, and air-gapped environments.
