JetStream's Clearance Puts A Zero-Trust Brake On Every AI Agent Action

JetStream's new Clearance reasoning engine authorizes every AI agent MCP call in flight, backed by Redpoint and the CrowdStrike Falcon Fund and demoed at Fal.Con 2026.

JetStream's Clearance Puts A Zero-Trust Brake On Every AI Agent Action

JetStream Security has taken the wraps off Clearance, a reasoning engine that authorizes every AI agent tool call before it executes rather than after, moving the industry's zero-trust conversation from the identity boundary all the way down to the individual action an agent is about to take. The Santa Clara startup announced the product on September 2 and is demoing it at Fal.Con 2026 in Las Vegas.

Gateways Route Traffic. Clearance Decides Whether The Traffic Should Have Left In The First Place.

AI gateways have quietly become commodity infrastructure over the summer, with Rippling, Ramp, Databricks, Microsoft and IBM all shipping one and joining Portkey, Kong, LiteLLM and JetStream itself. Anthropic already recommends putting a gateway in front of enterprise Claude Code deployments. But those gateways cache, rate-limit and log; none of them evaluate whether a specific action ought to run. Clearance sits at the JetStream AI Gateway and, on every request, checks the agent identity, the approved Blueprint, the tool being invoked and what that call is about to do. Sequences that resemble known exfiltration patterns - for example, a legitimate query, attachment prep and email send that suddenly acquires an external BCC - are denied clearance before the send.

Backed By Redpoint And CrowdStrike, Aimed At Fleets Of Agents

JetStream is backed by Redpoint Ventures and the CrowdStrike Falcon Fund, and Clearance ships alongside JetStream's inclusion in the new CrowdStrike AI Partner Specialization launched at Fal.Con. Chief executive Raj Rajamani framed the target market as enterprises that already have tens of thousands of employees and are preparing to run hundreds of thousands of AI agents under company-issued identities.

CrowdStrike logo

Bound To Blueprints, Enforced At The Parameter Level

Clearance rides on JetStream's AI Blueprints - the versioned operational contracts that describe every model, MCP server, tool, dataset and identity in an agent stack. Policy binds to the identity at access time and reaches inside the tool down to individual parameters, so an operator can enable a server's read tool while blocking its delete tool. Clearance enters general availability this fall.

Related coverage: CrowdStrike SafeMind with NVIDIA Nemotron, AIR's $50M firewall for AI agents, and Huskeys' $27M for agentic AI traffic control.

Reporting based on coverage from ACCESS Newswire and JetStream Security.

Category: Cyber Security

Tags: AI Startups AI Cybersecurity AI Agents Anthropic AI Security

Related Articles