Skip to content
AgentsIndex

Lasso

AI security platform tying agent discovery, automated red teaming and CPU-based runtime guardrails into one loop.

Lasso is an enterprise AI security platform that discovers the AI agents, applications and MCP servers in use across a company and assesses their risk. It red-teams them with automated attacks and enforces policies inline at the proxy, API or AI gateway layer.

Best for
Lasso Security is best for security teams who need to govern agents, apps, and MCPs across the AI lifecycle.
Look elsewhere if
  • you need the full API on every plan

What we checked

Pages last read
Facts checked
13 facts

From 12 sources

AI answers
Too few read

58 buyer-question answers read, November 2026

At a glance

Tags
ISO 27001, SOC 2

For developers

Works with
CI, AI Gateway, Vertex AI, Microsoft Copilot, AWS Bedrock, Salesforce Agentforce and 5 more

Lasso in our research

All research

Study

AI Tool Pricing Index, November 2026

Pricing

Contact sales

82 of the 361 tools we checked ask you to contact sales for a price.

Edition price, checked 10 November 2026.

Pricing pages of the AI tools, November 2026 editionas of n = 361 tools checked

A security layer that sits in the path of agent and LLM traffic

Lasso is an enterprise platform for securing AI agents and the applications built on models, sold as one product with five stages. Its platform page lists discovery and an AI-BOM of every agent, its models, system prompts, tools and guardrails; posture management against NIST and OWASP; automated red teaming with multi-turn agentic attacks; runtime enforcement; and detection and response mapped to MITRE and OWASP.

At runtime it works as a proxy rather than a code scanner. Lasso says it enforces policy "at the proxy, API, or AI Gateway layer", and its gateway page names Kong, Envoy AI Gateway, TrueFoundry, Portkey and LiteLLM as gateways it plugs into, inspecting user-to-agent, agent-to-agent, LLM, MCP and tool traffic. On a violation it blocks, masks or alerts, which the vendor says happens "in under 50ms with no manual intervention required". Discovery connects to CI/CD pipelines and to agent builders, with Vertex AI, Bedrock and Salesforce named. A separate AI Usage Control line covers employee use of outside AI tools through a browser extension and proxy integrations.

No rate card: a demo, a free assessment or the open-source gateway

Lasso publishes no price for the platform. There is no self-serve sign-up, no trial of the platform and no stated contract term or cancellation policy.

The free assessment is a one-off engagement, not a free plan: a "Targeted AI Red-Team Assessment" in which Lasso runs autonomous multi-turn attacks against one application and returns an executive and technical summary, reproducible exploit paths and remediation guidance. It is requested through a form and functions as the start of a sales conversation.

The only part a team can use at no cost without speaking to anyone is the open-source MCP Gateway, covered below.

The MIT-licensed MCP Gateway is the piece you can run today

It runs as an MCP server itself, reading an existing Cursor or Claude config and proxying the servers listed there through two tools, get_metadata and run_tool, while sanitizing requests and responses.

Guardrails come as plugins. The marketing site shows no such sign-up, so the free path stops at secret and PII masking.

Lasso Security Inc. raised $30 million in September 2026

The privacy policy names the operator as Lasso Security Inc. A press release dated September 2026, datelined Tel Aviv and New York, says the company was founded in 2023 and announced "a $30 million funding round led by ClearSky, with participation from Entrée Capital (which led Lasso's earlier seed round and increased its position in this round)", alongside iAngels, Singtel Innov8, Mindset and Swish Data. Elad Schulman is quoted as Co-Founder and CEO and Ophir Dror as Co-Founder and CPO.

The same release launched two guardrail engines. LEAP is described as transformer-free, running on ordinary CPUs at "under five milliseconds per decision" and deployable in "the customer's own cloud, regulated environments or air-gapped networks". RAPID is "a self-hosted LLM-as-a-judge" for the harder policy calls. The release names the US Department of Homeland Security, Kaufman Rossin, eToro and Fiverr as customers and claims revenue grew "by over 500%" in 12 months. The home page cites placements in several Gartner reports and a Leader rating in Latio's 2026 AI Security Market Report.

SOC 2 Type II is its own audit; other frameworks are mappings

The only audit Lasso states for itself is SOC 2. The September 2026 release says Lasso "is SOC 2 Type II certified (Security criteria)".

Product documentation is closed: docs.lasso.security states "Lasso documentation is only available to signed-in users" and offers separate US and EU logins, so API references and deployment guides cannot be read before buying. The privacy policy keeps data "until we delete our records" or until a customer sends a valid deletion request, and describes no configurable retention period. Nothing public covers SSO, an uptime SLA or whether customer prompts train Lasso's models.

Its headline figures are self-reported and do not all agree

The numbers Lasso leads with come from its own testing. The platform page claims "98.6%" detection accuracy and calls Lasso "570x" more cost-effective than cloud-native guardrails, and the LEAP release rests its accuracy claim on "Lasso's head-to-head benchmarks". No independent test of these figures is linked.

The same platform page gives two sizes for the red-team attack library as of November 2026.

Where it fits, and where it does not

Lasso suits a security team at a large organization that already runs agents in production, routes model traffic through a gateway such as Kong, Portkey or LiteLLM, and wants discovery, red teaming and inline enforcement from one vendor, including deployments in its own cloud or an air-gapped network. Its stated customers are enterprises and US federal agencies.

It fits poorly for a small team or a solo developer. There is no published price, no self-serve plan and no public documentation, so cost and integration effort are only known after a sales process. A developer who wants basic MCP hygiene can run the open-source gateway's secret and PII masking for free, but prompt-injection detection and custom policies need the paid Lasso API behind it.

Lasso price and terms

Terms

Free to start
PartlyLimited free useNo free plan of the platform
The terms in full

Lasso offers a one-off free targeted AI red-team assessment via a form, and its MCP Gateway is free open-source software under the MIT license.

For AI agents

Developer docs

Where these prices come from

lasso.security/book-a-demo, checked

What we checked on Lasso

YesPartlyNo

What it costs to startYes

Free to use

Only the open-source MCP Gateway is free to install and run; the commercial Lasso platform is sold through demos and its premium guardrail plugin needs a Lasso API key.

Check out Lasso's open-source MCP gateway

lasso.security/use-cases/mcp-securityChecked

What it works withYes

Works with

Discovery connects to CI/CD pipelines, cloud platforms and agent builders such as Vertex AI, Bedrock and Salesforce; runtime works with AI gateways including Kong, Envoy, TrueFoundry, Portkey and LiteLLM. No plan gating is published.

Lasso connects directly to your CI/CD pipelines and existing cloud platforms or third-party agent builder tools, including Vertex AI, Bedrock, and…

lasso.security/platform/ai-securityChecked

What it works withPartly

Public API

Lasso enforces policy at an API layer and its MCP Gateway calls Lasso's API with a key, but the product documentation is only available to signed-in customers, so there is no public API reference.

Lasso documentation is only available to signed-in users.

docs.lasso.securityChecked

What it does on its ownYes

Acts on its own

Inline enforcement blocks, masks or alerts on violations on its own at the gateway layer, and the red-teaming module runs autonomous multi-turn attacks without a person approving each step.

All enforcement runs in under 50ms with no manual intervention required.

lasso.security/use-cases/ai-gateway-sec…Checked

What it does on its ownYes

Limits on what it may do

Administrators choose what the enforcement does on a match (block, alert or sanitize) and set role-based permissions and DLP policies, so its actions are configured in advance.

Configurable actions: block, alert, or sanitize based on your policies

lasso.security/use-cases/mcp-securityChecked

What it does on its ownYes

No code needed

AI Usage Control deploys through a browser extension and proxy or data integrations, with plug-and-play policy templates; securing homegrown apps uses proxy, SDK or API-level integration instead.

Enable full AI usage control in minutes through Lasso’s lightweight deployment paths like our browser extension, proxy integrations, and data…

lasso.security/platform/ai-usage-controlChecked

What happens to your dataPartly

Retention you control

The privacy policy keeps data until Lasso deletes its records or a customer sends a valid deletion request, with longer retention for legal or accounting needs; no configurable retention setting is described.

Your data will be stored until we delete our records, and we proactively delete it, or if you or our Customers send a valid deletion request.

lasso.security/privacy-policyChecked

What happens to your dataYes

Data stays in your region

The LEAP guardrail engine can be deployed in the customer's own cloud, regulated environments or air-gapped networks, keeping inspection where the data must stay; residency for the hosted platform is not described.

LEAP deploys anywhere the customer’s data has to stay, be it the customer’s own cloud, regulated environments or air-gapped networks.

lasso.security/resources/lasso-security…Checked

What happens to your dataYes

Self-hosted

The LEAP guardrail runs on ordinary CPUs inside customer environments, and its companion RAPID engine is described as a self-hosted LLM-as-a-judge; the open-source MCP Gateway also runs locally.

RAPID is a self-hosted LLM-as-a-judge that handles the small fraction of decisions requiring complex, plain-language policy judgment

lasso.security/resources/lasso-security…Checked

What happens to your dataYes

Open source

The MCP Gateway is published on GitHub under the MIT license; the commercial platform, including its advanced guardrail plugin, is not presented as open source.

On the page: Repository sidebar, License

github.com/lasso-security/mcp-gatewayChecked

Whether a company can buy itYes

Security audits

Lasso announced its own SOC 2 Type 2 compliance in February 2024; ISO 27001, ISO 42001 and HIPAA appear on its pages as frameworks it maps customers to, not as its own certifications.

As of February 2024, Lasso Security is proud to announce that we are officially SOC 2 Type 2 compliant.

lasso.security/blog/lasso-security-cert…Checked

Whether a company can buy itYes

Audit log

Every AI interaction is captured with its execution context and mapped to compliance frameworks, and the vendor says these audit trails can be exported.

Audit trails are exportable and current, not generated retrospectively.

lasso.security/use-cases/ai-gateway-sec…Checked

Questions about Lasso

How much does Lasso cost?
Lasso gives its price on request. From lasso.security/book-a-demo, checked 10 November 2026.
Does Lasso have an API or an MCP server?
Lasso has a limited API. From docs.lasso.security, checked 10 November 2026. Lasso has an MCP server. We checked mcp.lasso.security/.well-known/oauth-protected-resource ourselves on 15 November 2026.
Is Lasso open source, and can you host it yourself?
Lasso is open source. From github.com/lasso-security/mcp-gateway, checked 10 November 2026. You can host Lasso yourself. From lasso.security/resources/lasso-security-announces-future-of-ai-security-with-cpu-based-guardrails, checked 10 November 2026.
What have we checked on Lasso?
13 facts, under these headings: What it costs to start, what it works with, what it does on its own, what happens to your data, whether a company can buy it.
How do we check Lasso?
We hold 13 facts on Lasso, quoted from 12 pages. Facts last checked 15 November 2026. Price last checked 10 November 2026. Last change found 15 November 2026.

How we check every listing

Lasso price history

We have read the price once so far, on 10 November 2026: price on request.

PricePlanFirst seenLast seenReadingsSource
Price on request10 November 202610 November 20261Pricing page, read 10 November 2026

From our research on 10 November 2026. A date is when we read the page, not when the price changed.

Similar to Lasso

Is Lasso yours?

Free, and the page then carries the verified owner mark. You confirm it with an email at lasso.security. Claim this listing

Report an error if a fact is wrong.