# Scalekit
An AI agent integrations platform for connecting agents to tools with MCP gateways, scoped access, and per-user auth.
Listing: https://bestaithings.com/tool/scalekit

Website: https://www.scalekit.com/
Categories: Automation
Pricing: Freemium, from $99/month
Runs on: Web
Public API: yes
MCP server: yes
License: proprietary
Listing updated: 2026-11-17

Scalekit's AgentKit puts one MCP gateway between an AI agent and the tools it calls. It covers 600+ connectors and 30,000+ tools across CRM, email, calendar, tickets, docs, and billing, so you build the agent rather than the integration layer.

Users connect their own accounts. The gateway stores each user's tokens, encrypts them per tenant, refreshes them before they expire, and attaches them to every call, so your code and prompts never hold a token. The agent acts as the user it is working for, with the permissions that user granted, and one user's run cannot reach another's. Auth methods include OAuth 2.0, API keys, and bearer tokens, behind a consent screen on your own domain. The gateway works with agent frameworks including Anthropic, OpenAI, LangChain, Google ADK, Mastra, and Vercel AI, and setup starts with an npx CLI command.

Every call is logged with the agent, user, tool, and result. Logs are searchable, exportable, and can stream to Datadog, Splunk, or any SIEM. Failures are marked as connector errors or API errors.

### What you can do with it

- **Connect agents to tools.** Call 30,000+ tools across 600+ connectors through one MCP endpoint, each written so a model can pick the right one and send the right arguments.
- **Scope tools per agent.** Virtual MCP servers give each agent only the tools its role needs, picked from any app and served as one endpoint. This cuts tool overhead in the agent's context.
- **Start runs on events.** Triggers fire an agent when something happens in a connected account, such as a new email, a deal changing stage, or a failed payment, with no polling, cron jobs, or webhook endpoints to run.
- **Add your own tools.** Extend a connector with an endpoint it lacks, add a new system using OAuth 2.0, API keys, or bearer or basic auth, or bring your own MCP server. These get the same auth, scoping, and logs.
- **Start from agent templates.** Open-source templates cover jobs like security vulnerability triage, release changelogs, new hire provisioning, incident response, and revenue forecast commentary.

### Who it's for

Developers and engineering teams shipping AI agents that act on behalf of their customers' accounts in tools like Salesforce, Gong, Slack, and Google Drive, and who do not want to build OAuth handling and token storage themselves. Enterprise options include bring-your-own-key encryption with AWS or GCP KMS, and VPC or self-hosted deployment. It is a developer product, not an end-user app.

## Questions
Q: Is Scalekit free?
A: Yes, Scalekit has a free plan. Paid plans start at $99/month.
Q: How does Scalekit handle user credentials for agents?
A: Each user connects their own accounts, and the gateway stores the tokens, encrypts them per tenant, and refreshes them before they expire. Your code and prompts never hold a token, and one user's run cannot reach another's.
Q: Can Scalekit send tool call logs to my SIEM?
A: Yes, logs can stream to Datadog, Splunk, or any SIEM as calls happen. Each event carries the agent, user, tool, and result, and failures are marked as connector or API errors.
Q: Can I add tools that Scalekit's catalog doesn't include?
A: Yes, you can add internal services, partner APIs, or endpoints a connector lacks, using OAuth 2.0, API keys, or bearer or basic auth. You can also bring your own MCP server, and its tools join the catalog with the same auth, scoping, and logs.
Q: Which agent frameworks does Scalekit work with?
A: It works with Anthropic, OpenAI, LangChain, Google ADK, Mastra, and Vercel AI, along with OpenClaw and Claude Code.
Q: Does Scalekit have an API?
A: Yes. Scalekit has a public API, and its documentation is linked at the top of this page.
Q: Does Scalekit have an MCP server?
A: Yes. Scalekit has an MCP server, so AI assistants such as Claude and ChatGPT can use it directly.
Q: What are some alternatives to Scalekit?
A: Other tools listed here for similar work include Composio, Glama, and Workato.
