Lasso Security is an AI security vendor focused on agents and applications. As of 17 October 2026, its site states the goal as "Protection for every AI agent you build and deploy, wherever it runs." Teams searching for Lasso Security alternatives are usually comparing agent-native security vendors, weighing a platform from a large security company, or finding that their main need is authorization of agent actions, not detection. This guide covers what Lasso offers, where it is strong, and how the categories compare. Other vendor guides are on the comparison hub.

What Lasso Security covers (as of 17 October 2026)

Lasso's site organizes the platform as discover, assess and protect:

  • Discover. Agent discovery through platform integrations, and discovery of homegrown applications through CI. The inventory maps models, system prompts, tools and guardrails, and tracks red-teaming scans and policies.
  • Posture. AI security posture management with misconfiguration and policy gap analysis, supply chain risk assessment, and alignment to NIST and OWASP frameworks.
  • Exposure assessment. Adversarial testing with what Lasso calls a 3,000+ attack library across the OWASP Top 10 and more, multi-turn agentic attacks, and tests for context poisoning and tool chain manipulation. Lasso says the results update policies automatically through purple teaming.
  • Protect. Inline policy enforcement "at the proxy, API, or AI Gateway layer", with detection mapped to MITRE and OWASP and intent analysis for behavioral anomaly detection.

Lasso publishes an Intent Security Framework white paper that argues AI security should analyze the intent behind an agent's actions, not only fixed patterns. Its site also announces CPU-based guardrails and lists analyst mentions, including a Leader placement in Latio's 2026 AI Security Market Report and a sample-vendor listing for AI Governance Technologies in Gartner's Hype Cycle for AI, 2026.

The open-source MCP gateway

Lasso maintains an open-source MCP Gateway under the MIT licence. Its README describes a plugin-based intermediary between LLMs and MCP servers. It reads server configuration from an mcp.json file, manages the lifecycle of the configured servers, intercepts requests and responses to sanitize sensitive information, and includes a scanner that checks server reputation before loading a server. Our MCP gateway comparison places it next to other MCP gateways; for gateways in front of model providers instead, see the AI gateway guide.

Where Lasso is strong

  • Agent-first scope. Discovery, testing and runtime protection all start from the agent and its tools, not from a network or model-file view.
  • Testing that feeds policy. Adversarial results that update runtime policies close a loop many teams run by hand.
  • Flexible enforcement points. Proxy, API and gateway options let a team place protection where its architecture allows.
  • Open-source entry point. The MCP gateway lets developers try sanitization on MCP traffic without a contract.

Why teams evaluate alternatives

  1. A platform vendor already in place. A company that runs Palo Alto Networks or Cisco may prefer that vendor's AI security product for procurement reasons.
  2. A narrow need. A team adding guardrails to one application may want a guardrail API and nothing else.
  3. Authorization and approvals. Threat detection and enforcement, even with intent analysis, focus on whether an action looks risky. Some teams also need an explicit policy that allows or denies each tool call for each agent, with a person approving some of them.
  4. Evidence an auditor checks offline. Some buyers want exported decision records they can verify without the vendor's console.

Alternatives by category

What you need Category Examples covered on this site
Agent-focused discovery, testing and runtime protection Agent-native security Lasso, Noma Security, Zenity, Prompt Security
AI security inside a wider security platform AI security platform Prisma AIRS, Cisco AI Defense
Prompt and response guardrails through an API Guardrail API Pangea, Lakera
Agent identity, tool-call policy, approvals and signed evidence Agent control plane Praesidia

Our post on platform-bundled versus pure-play agent security explains how buyers usually choose between the first two rows.

Where Praesidia fits

Praesidia is our product. It is an AI agent control plane. Its focus is narrower than Lasso's on testing and detection, and different on authorization and evidence. What it does:

  • Identity per agent. Each registered app, agent and MCP server gets its own credentials for authentication and attribution.
  • Tool access rules. Rules are set per agent and scoped to one MCP server or all of them. Without a matching allow rule, a governed tool call is denied.
  • Approvals. A step-up rule turns an MCP tool call into a durable approval request that expires if nobody answers.
  • Guardrails in the model gateway. Rule matches can block, warn, redact or replace; ML and LLM guardrails block or warn.
  • Signed record. Praesidia records the decisions it makes in an append-only, hash-chained audit trail, signed under the default configuration; an operator can turn signing off. A signed record shows it was not changed afterwards. It does not show that every action was captured. Calls that do not pass through Praesidia are not recorded.

On testing, Praesidia's red-team probes check organization-wide input guardrails and intent rules; they do not call the production target or test its agent-specific guardrails. A team that needs deep adversarial testing of its own agents should weigh that. See AI agent runtime security for how the per-call decision works.

Questions to ask in an evaluation

  1. How does discovery reach your agents? Platform integrations, CI and traffic each find different agents. Ask which of yours each one finds.
  2. What happens on a detection? Ask whether a detected threat blocks the call in band, and what the agent sees.
  3. Can policy decide on tool arguments? A rule that reads the tool name and arguments is different from one that reads only text.
  4. Who approves a risky action, and where is that recorded?
  5. Open source or platform? The MCP gateway is MIT-licensed; confirm which features need the commercial platform.

Common questions

What is Lasso Security? Lasso Security is an AI security vendor. As of 17 October 2026, its platform covers discovery of AI agents and applications, posture management, adversarial testing and inline runtime protection.

Is Lasso's MCP gateway open source? Yes. The lasso-security/mcp-gateway repository is MIT-licensed. It proxies MCP servers listed in an mcp.json file and sanitizes sensitive information in requests and responses through plugins.

What is the difference between Lasso and an agent control plane? Lasso's public focus is finding agents, testing them, and detecting and enforcing policy on threats at runtime. A control plane gives agents their own identities, decides governed tool calls against policy and holds some calls for a person. Many teams need parts of both.

Does Praesidia replace Lasso? Not for adversarial testing or broad discovery. Praesidia governs the agents and MCP tool calls routed through it and produces signed evidence of its decisions.

Sources (as of 17 October 2026)

  • Lasso Security, lasso.security: platform capabilities, Intent Security Framework, CPU-based guardrails announcement, analyst mentions.
  • Lasso Security on GitHub, mcp-gateway: README and MIT licence.