Kindo Raises $30 Million to Control Enterprise AI Agents

By Grace Kim

Kindo has raised $30 million in a financing expansion, bringing total capital raised to $57.6 million, the Venice Beach, California-based company said.

Kindo said Crosspoint Capital and Forgepoint Capital co-led the round. The company will use the proceeds to expand its go-to-market organization and accelerate development of its AI Control Center, Deep Hat models and add-ons. Kindo reports approximately 800% year-over-year revenue growth in 2026.

Andre Fuetsch, managing director at Crosspoint Capital, and Ernie Bio, managing director at Forgepoint Capital, will join Kindo’s board of directors, according to Kindo.

What problem is Kindo trying to solve?

Deploying an agent delegates authority. An agent linked to business systems may read internal information, update records, approve requests or modify infrastructure. Kindo says security and operations teams need visibility into both the permissions agents hold and the actions they take as deployments multiply.

Kindo positions its platform as a shared operating layer for IT, security, finance and compliance. Its discovery maps sanctioned and shadow AI, while a central view brings together activity, access, costs and outcomes. Organizations can build agents within Kindo or connect agents developed elsewhere.

Kindo calls this an “AI Operations Center,” borrowing continuous monitoring and intervention from security and network operations centers. Its registry records agent ownership, permissions inherited from the person running an agent and cost attribution. Those functions identify responsibility, control authority and test whether an agent’s work justifies its resource use.

Founder and CEO Ron Williams said: “The gap is not a shortage of agent builders. It is the tooling organizations need to control how agents are built, which data they can reach, where information can go, and what operating those agents costs.”

In practical terms, Kindo wants one place to identify who owns each agent, what it can access and what it costs.

How do gateways differ from the agent harness?

According to Kindo’s announcement, its gateways govern and audit agent calls to connected data, tools and approved models. Identity controls apply existing organizational permissions to people and agents, with the option to tighten access centrally. A policy layer governs access across supported integrations, though coverage depends on the agents and resources connected to it.

For mission-critical deployments, Kindo offers a purpose-built agent harness that manages an agent’s execution. Operators gain monitoring, support and intervention capabilities, including taking over an agent when something goes wrong. The funding release associates deeper intervention and several Deep Hat cost features with agents running inside the harness.

Connecting an externally built agent through a gateway differs from running it inside Kindo’s execution environment. Kindo says the distinction matters commercially and technically because deeper intervention and several Deep Hat cost features apply to harness-based agents.

The gateway sets permissions and audits calls; the harness controls the run.

How does Kindo measure agent spending?

The AI Control Center records token consumption and agent-run outcomes, including failed runs, Kindo said. It attributes spending to individuals, agents, departments and projects. Finance teams can set budgets, receive alerts and enforce hard spending caps.

Kindo says cost accounting should be read alongside outcomes, not model price alone. A low-priced agent that repeatedly fails can waste resources, while an expensive model may be justified for a difficult, valuable task. Linking cost and outcome can expose repeated failures, unnecessary model capacity and work that could be handled more efficiently.

Kindo also offers four deployment choices: its cloud, a self-managed customer cloud, on-premises infrastructure or an air-gapped environment. The company says deployment location and model selection remain separate decisions, though both affect control, operational requirements and cost.

Kindo ties each agent’s bill to its results and then offers controls over where the software runs.

Which costs do the Deep Hat tools target?

Kindo’s funding release describes four Deep Hat tools:

• Sift: Learns which events an agent would dismiss, filters them after a person approves the threshold, records filtered events and sends a sample onward for continued measurement.
• Tune: Adapts a smaller, lower-cost model to a particular task and reserves larger models for work that needs them.
• Shrink: Turns repeatable workflow steps into code, removes the model cost of those steps and routes exceptions back to a model.
• Attack: Supports private, authorized offensive security testing instead of using frontier cybersecurity models for routine testing.

Kindo’s website says Sift starts in shadow mode, allowing every event to reach the target agent while it learns from that agent’s verdicts. A person must approve activation, and a sample of events that would be filtered still reaches the agent to check for misses.

Kindo says Sift measures agreement against the target agent’s own verdicts. That agreement should not be confused with independently verified ground truth.

The four tools attack distinct cost sources, but Sift still depends on the target agent’s judgments.

Who uses Kindo, and who backed the round?

Kindo says its platform runs in production at large and mid-market enterprises across financial services, professional services, healthcare, and aviation and transportation.

The announcement highlights Aireon. Aireon’s CISO reported in May 2026 that using Kindo for threat hunting and Tier 1 and Tier 2 security operations helped generate more than $2 million in annual savings. The release presents that as a customer-reported outcome, not a guaranteed result for other deployments.

Kindo named TFX Capital and CircleRock Capital as participants alongside existing investors Drive Capital, Riot Ventures, RRE Ventures, Eniac Ventures, Marlinspike Partners, New Era Ventures and Sunset Venture.

The capital gives Kindo room to scale go-to-market work and accelerate AI Control Center and Deep Hat development; its next test is whether the governance, cost controls and cited savings hold across deployments beyond the company’s announcement.