Klavis AI

Klavis runs hosted MCP servers for more than a hundred applications, from Affinity and Airtable to Zendesk and Zoho Mail in its docs catalog, and connects each one per end user through OAuth 2.0 or an API key, so an agent reads and writes in that user's own accounts. Tools are reached through each server's MCP URL, through Strata (one MCP server that lets the agent discover categories and actions across every connected server before executing one), or through the Call Tool API for plain function calling.

Sourceklavis.ai/docs/mcp-server/overviewread 2026-09-22

The product executes the tool calls an agent makes and returns the results. Sequencing, branching and handoff live in the customer's own framework, for which Klavis publishes guides (LangChain and LangGraph, LlamaIndex, CrewAI, Agno, Mastra, Google ADK and the major model APIs). Strata's discover, browse and execute tools help the agent pick an action, which is tool selection rather than orchestration. No workflow runtime of Klavis's own is documented to sequence, branch, retry or route work, or to combine deterministic workflow nodes with autonomous agent steps.

Sourceklavis.ai/docs/ai-platform-integration/overviewread 2026-09-22

There is no retrieval structure over the customer's own content. Klavis's MCP servers for Notion, Google Drive, Confluence and similar tools let an agent query those systems live, which is integration rather than grounding. Strata finds tools rather than documents, by browsing categories and actions instead of semantic search. No ingestion, indexing, retrieval or RAG layer of the product's own is documented, and no retrieved passages or sources are shown.

Sourceklavis.ai/docs/concepts/strataread 2026-09-22

The agent works inside constraints. Each end user authorizes an integration through OAuth consent before its tools act, and a server connection can be set read only through the API. Klavis Guardrails is described as a proxy between MCP clients and servers that detects tool poisoning and prompt injection, enforces least privilege access and checks tool invocations against allowlists. No approval step a person acts on before an action commits is documented, whether at a node, workflow or policy layer.

Sourceklavis.ai/docs/enterprise-security/klavis-securityread 2026-09-22

The API enforces HTTPS with API key authentication, each end user's credentials are scoped per integration and returned only to the key owner that controls the instance, and customers can run OAuth through their own white label OAuth apps. Connections can be set read only, and Klavis Guardrails enforces least privilege tool access. No attestation or certification is published: the FAQ's security answer names built in OAuth only, and the site footer links terms, privacy and cookie policies with no trust center. SSO, roles and an audit trail are not documented either.

Sourceklavis.ai/docs/api-reference/overviewread 2026-09-22

No record of what an agent did through Klavis is documented for the customer: the API returns server instances, tools, users and authentication status, and the docs send users to the dashboard to check authentication, but no log, trace, execution history or audit event of tool calls is documented, and no export to a SIEM. Guardrails' real time threat detection is enforcement, not an inspectable record.

Sourceklavis.ai/docs/api-reference/overviewread 2026-09-22

Each end user's credentials and integration status are stored so tools act without asking for consent again, and a sandbox holds application state for one training or evaluation run, but no memory layer that keeps an agent's context across runs is documented; Klavis's Mem0 server connects an agent to another vendor's memory product. Stored credentials and a test environment's data are not agent memory, and no memory scope or lifetime is stated.

Sourceklavis.ai/docs/api-reference/user/get-user-integrationsread 2026-09-22

Hosted servers run on Klavis's infrastructure with a stated 99.9% uptime SLA, and the API specification names two production servers, api.klavis.ai for the US and api.eu.klavis.ai for the EU. For on premises or custom deployments every MCP server ships as a Docker image to run in the customer's own infrastructure, and Strata is open source. Buyers can pick the US or EU region, or run the servers in their own environment.

Sourceklavis.ai/docs/concepts/mcpread 2026-09-22

The catalog holds MCP servers and tools, which give an agent actions rather than doing work when selected, and the framework guides build example agents as tutorial code. No ready made workflow, template or role specific agent of Klavis's own is documented for a buyer to adopt.

Sourceklavis.ai/docs/mcp-server/overviewread 2026-09-22

Slack and Discord clients come alongside the hosted servers, the FAQ says, so a person can put MCP tools to work from those chat channels. The documentation index carries no setup page for either client. Nothing that starts work without a person asking is documented: no schedule, webhook, event subscription or inbound queue.

Sourceklavis.ai/faqread 2026-09-22

Tools are served to whatever model the customer runs: the List Tools API returns a server's tools in several model formats, and the docs carry integration guides for OpenAI, Anthropic's Claude, Gemini, Mistral, Cohere, Together AI and Fireworks AI, each using the customer's own provider key. Klavis names no model of its own, so the model, the provider and the key are the customer's throughout.

Sourceklavis.ai/docs/ai-platform-integration/overviewread 2026-09-22

Klavis is callable from outside through a documented REST API at api.klavis.ai with a published OpenAPI specification, and Python and TypeScript SDKs cover servers, Strata, users, OAuth, white labeling and sandboxes. It serves MCP as an addressed service, each hosted server and each Strata instance having its own URL, and the servers and Strata are open source for self hosting.

Sourceklavis.ai/docs/installationread 2026-09-22

Isolated sandboxes of real applications let the customer test and train its agent. The Klavis Sandbox concept page describes a sandbox loaded with a deterministic world state, which the agent works against through MCP; the full state can be exported to compare against ground truth, and the sandbox resets for the next run. The Universe API acquires benchmark task environments, and its verify endpoint runs the task's check and returns whether the task passed, with the failure reason. Sandboxes can also be configured for the MCP-Atlas, Toolathlon and MCP-Mark benchmarks.

Sourceklavis.ai/docs/api-reference/universe/verify-taskread 2026-09-22

The API specification's Local Sandbox API starts a virtual machine hosting interconnected MCP servers, among them Playwright for a browser plus a terminal and Desktop Commander for the machine. Klavis describes it as serving filesystem, terminal, browser and document workflows during agent training and evaluation. Klavis restricts it to use outside production, and no Klavis page describes driving a browser through the hosted Playwright server any further. The hosted MCP servers and Strata work through application APIs, not interfaces.

Sourceklavis.ai/docs/api-reference/local-sandbox/acquireread 2026-09-22