Action-taking agent safeguardsAdditional if: AI takes multi-step actions · The employee-facing AI performs multiple steps or takes actions rather than only generating content.AI gateway and tool-connection controlsAdditional if: Model or tool traffic uses a security gateway · Calls to models, tools, or connections are routed through a security gateway, including Model Context Protocol (MCP) connections.AI data protectionDirectly addresses · Inspect prompts, responses, and files to prevent sensitive data from moving through AI tools.
Vendor research
Salt Agentic Security Platform
Review what this vendor says publicly, the security topics those statements may support, what remains unverified, and factual company context. This is not an assessment of product effectiveness or fit.
Use-case context
How this vendor relates to the selected use case
Related research availableBack to use case →
These links show approaches associated with this vendor. The relationship label describes how the approach maps to the use case—not product effectiveness, complete requirement coverage, or fit.
Company scale
Scaled?
ScaledA private provider with at least $100M in known funding or at least 250 employees.This is a company-scale signal, not a product-quality rating.
?
ScaledA private provider with at least $100M in known funding or at least 250 employees.This is a company-scale signal, not a product-quality rating.A descriptive band derived from retained public revenue, workforce, ownership, or funding signals.
Company scale is separate from product features, effectiveness, and suitability.
- $271M known funding
- Employee scale not sourced
- Private-company revenue and profitability not sourced
Company context
Private, VC-backed; no parent company or completed acquisition is identified on the reviewed Salt-controlled company and funding pages.
Scaled private vendor by the project's capital threshold; Salt currently positions its platform across agent discovery, posture management, runtime protection, Model Context Protocol (MCP) and agent-to-agent (A2A) security, application programming interface (API) protection, and AI-assisted development policy.
Research coverageCounts describe available public research, not product quality.View details
- Vendor statements
- 19 records
- Source-checked records
- 10
- Evaluation requirements
- 19 in this research model
- Unresolved requirements
- 9
Company intelligence
Who is behind the product
Company facts provide evaluation context. Each signal is kept separate because tenure, workforce, funding, and hiring answer different questions.
FoundedFirst-party source conflict: Salt's current About page says Roey Eliyahu and Michael Nicosia founded Salt in 2018, while Salt's 2022 Series D announcement says the company was founded in 2016. Preserve as unresolved pending vendor confirmation.
HeadquartersPalo Alto, California
OwnershipPrivate, VC-backed; no parent company or completed acquisition is identified on the reviewed Salt-controlled company and funding pages.
EmployeesNot yet sourced
Capital and scaleIndependent company
Salt Security
- Known funding
- $271M
- Operating scale
- Scaled private vendor by the project's capital threshold; Salt currently positions its platform across agent discovery, posture management, runtime protection, MCP and A2A security, API protection, and AI-assisted development policy.
- Backing context
- $140M Series D led by CapitalG at a stated $1.4B valuation; Salt reported $271M in total funding, with backing from CapitalG, Sequoia Capital, Y Combinator, Tenaya Capital, S Capital, Advent International, Alkeon Capital, and DFJ Growth.
Series D · $140M
Named investors
CapitalG · Sequoia Capital · Y Combinator · Tenaya Capital · S Capital · Advent International · Alkeon Capital · DFJ Growth
Founders and leadership2 people listed
- Roey EliyahuCurrent role listed
Co-Founder and CEO
- Michael NicosiaCurrent role listed
Co-Founder
Operating signalsRead each signal separately
There is no combined company rating. The company-scale label uses stated size thresholds; product features and effectiveness require separate evidence.
- Company tenure
- First-party source conflict: Salt's current About page says Roey Eliyahu and Michael Nicosia founded Salt in 2018, while Salt's 2022 Series D announcement says the company was founded in 2016. Preserve as unresolved pending vendor confirmation.
- Workforce scale
- Not yet sourced
- Hiring activity
- Not displayed
A current count requires a retained, clickable source URL.
Open research questions
- Current employee range is not yet supported by a public source.
Company sources and research limits5 linked public sources
Only company facts supported by retained public sources are shown. Missing values remain unknown, and company scale does not establish product effectiveness.
Company sourceSalt Agentic AI Security pageSalt describes continuous discovery of agent-related APIs, MCP servers, and LLM integrations, posture analysis, sensitive-data visibility, and real-time detection and prevention at the API action layer.Company sourceSalt AI-agent action security announcementSalt announced MCP Protect and Agentic AI Governance for visibility, guardrails, and real-time protection across AI-agent actions involving MCP and A2A.Company sourceSalt Code launch announcementSalt describes Salt Code as a component that applies policy during AI-assisted code generation, pipeline validation, and runtime, and states its launch availability and customer packaging.Company sourceSalt About pageSalt's current About page names its founders and current leadership and states a 2018 founding year.Company sourceSalt Series D announcementSalt's Series D announcement reports a $140M round, $271M total funding, a $1.4B valuation, participating investors, and a conflicting 2016 founding year.
Solution areas
These areas describe how the vendor approaches enterprise AI security. They do not establish product quality or fit.
AI API and web application protectionCore product focusAction-taking agent safeguardsCore product focusAI gateway and tool-connection controlsRelated coverageAI application runtime protectionRelated coverageCoding-agent and developer workstation securityRelated coverageAI data protectionRelated coverage
Buyer context
- Treat Salt as an agentic-security and application programming interface (API) action-layer candidate because its current first-party material explicitly covers AI agents, Model Context Protocol (MCP) servers, agent-to-agent (A2A) interactions, and runtime protection.
- Salt's strongest public evidence concerns discovery, graph context, application programming interface (API) and Model Context Protocol (MCP) posture, sensitive-data flow, behavioral detection, and runtime guardrails; validate how blocking is implemented in the buyer's environment before describing it as an inline gateway.
- Salt Code extends the platform into AI-assisted development policy, but reviewed public material does not establish broad workstation command, filesystem, network, secret, hook, extension, and package enforcement.
- The two conflicting first-party founding years should remain visible in the record until Salt confirms the correct corporate founding date.
Related frameworks
Where public vendor statements relate to framework requirements
11 related frameworks · expand when needed
Related frameworks
Where public vendor statements relate to framework requirements
These links show related requirements for further review. They do not establish framework compliance or control implementation. Open the full framework crosswalk →
Current referenceCSA AI Controls Matrix
- Requirements with public support
- 9
- Related requirements
- 14
- References
- 71
- Requirements with public support
- 8
- Related requirements
- 11
- References
- 58
- Requirements with public support
- 8
- Related requirements
- 12
- References
- 35
- Requirements with public support
- 8
- Related requirements
- 12
- References
- 71
- Requirements with public support
- 8
- Related requirements
- 13
- References
- 39
- Requirements with public support
- 8
- Related requirements
- 12
- References
- 39
- Requirements with public support
- 8
- Related requirements
- 12
- References
- 31
- Requirements with public support
- 8
- Related requirements
- 11
- References
- 27
- Requirements with public support
- 8
- Related requirements
- 13
- References
- 31
- Requirements with public support
- 8
- Related requirements
- 13
- References
- 25
- Requirements with public support
- 3
- Related requirements
- 3
- References
- 3
Evaluation questions
What to verify beyond public claims
These questions come from security requirements with some public support. Use them as starting points for demonstrations, documentation review, customer references, or a buyer-observed pilot.
- 01Approved AI usage monitoring
Approved AI workspace activity appears with user, workspace or tenant, model or provider, and timestamp.
- 02Approved AI usage monitoring
Prompt, model, or admin activity can be exported or correlated for the selected approved AI platform.
- 03Sensitive-data protection for generative AI
Sensitive prompt, response, or file test data is detected and classified during an AI interaction.
- 04Sensitive-data protection for generative AI
A policy redacts, blocks, coaches, or records the sensitive data event before it leaves the approved path.
- 05Generative AI application security
A test large language model (LLM) application event records prompt, application programming interface (API), model, retrieval, or tool interaction context.
- 06Generative AI application security
A prompt-injection or unsafe-output test is detected, blocked, or flagged by the guardrail or large language model (LLM) firewall.
- 07AI gateway, tool-connection, and runtime controls
A model, agent, tool, or Model Context Protocol (MCP) request passes through a named policy enforcement point.
- 08AI gateway, tool-connection, and runtime controls
A test policy allows, blocks, transforms, redirects, or rate-limits the request with an explicit reason.
Detailed security-requirement research18 evaluation items · supporting evidence and open research are shown separatelyExpand
Security requirementPublic supportRelated frameworksWhat to verify
Unapproved AI use discovery
No supporting claim foundDiscover and monitor workforce AI tools, accounts, prompts, domains, models, users, and usage outside approved controls.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
An unmanaged AI app used by a test user appears in discovery inventory with user, app or domain, and timestamp.
AI-feature discovery in business applications
No supporting claim foundInventory software as a service (SaaS) applications that embed AI features, expose enterprise data to AI capabilities, or create AI-driven data movement.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A software as a service (SaaS) app with an embedded AI feature appears in the software as a service (SaaS) AI inventory with app, provider, and feature context.
Approved AI usage monitoring
Limited public supportMonitor approved AI workspaces, tenants, gateways, and model platforms such as ChatGPT Enterprise, Claude Enterprise, Gemini, Microsoft Copilot, Vertex AI, Elvex, or internal AI gateways.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
Approved AI workspace activity appears with user, workspace or tenant, model or provider, and timestamp.
Controls for unapproved AI use
No supporting claim foundBlock, coach, redirect, or contain non-approved AI use and policy-violating AI interactions.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A policy blocks, coaches, redirects, or contains a test interaction with an unapproved AI destination.
Sensitive-data protection for generative AI
Strong public supportDetect, classify, redact, or block sensitive data in prompts, responses, files, retrieval, memory, and AI-connected workflows.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
Sensitive prompt, response, or file test data is detected and classified during an AI interaction.
Browser and business-application controls
No supporting claim foundApply session-level controls in browser and software as a service (SaaS) workflows, including uploads, downloads, copy/paste, sharing, and identity-aware access decisions.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A session-level policy controls upload, download, copy, paste, sharing, or form submission in a browser or software as a service (SaaS) workflow.
Generative AI application security
Limited public supportProtect enterprise-built large language model (LLM) applications, retrieval-augmented generation (RAG) systems, prompts, application programming interfaces (APIs), model calls, tools, and production runtime behavior.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A test large language model (LLM) application event records prompt, application programming interface (API), model, retrieval, or tool interaction context.
AI governance, risk, and compliance
No supporting claim foundInventory AI systems and owners, translate policy and regulatory obligations into governed workflows, assess risk, manage approvals and exceptions, and retain audit evidence across the AI lifecycle.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A test AI system is registered with owner, intended use, risk tier, lifecycle state, and applicable obligations.
AI assurance and adversarial testing
No supporting claim foundTest models, applications, retrieval-augmented generation (RAG) systems, and agents before release and continuously thereafter using adversarial probes, evaluation suites, attack simulation, and security release gates.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A controlled test campaign exercises an AI model, application, or agent against named AI attack classes.
AI model and supply-chain security
No supporting claim foundDiscover, inventory, scan, validate, and monitor models, datasets, model artifacts, registries, dependencies, and AI development assets for tampering, unsafe serialization, provenance gaps, or malicious content.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A test model or AI artifact appears in inventory with origin, version, hash or provenance, and deployment context.
AI gateway, tool-connection, and runtime controls
Limited public supportMediate model, agent, tool, application programming interface (API), connector, and Model Context Protocol (MCP) traffic through an enforcement point that applies identity-aware policy, content controls, routing, rate limits, and auditable allow or deny decisions.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A model, agent, tool, or Model Context Protocol (MCP) request passes through a named policy enforcement point.
Action-taking agent monitoring
Strong public supportObserve and govern agent plans, memory, tool calls, delegated tasks, autonomy, runtime decisions, and outcomes.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A test agent run captures plan, steps, tool calls, outcome, and timestamps.
Agent-to-agent communication security
Limited public supportAuthorize, log, and control agent-to-agent, agent-to-tool, Model Context Protocol (MCP), connector, and tool-chain handoffs.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
An agent, tool, connector, or Model Context Protocol (MCP) handoff logs source identity, destination, and authorization decision.
Non-human identity and service-account security
No supporting claim foundInventory, least privilege, credential hygiene, monitoring, and lifecycle management for non-human identities, workloads, service accounts, application programming interface (API) keys, and machine credentials.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A test service account, agent identity, or non-human identity appears in inventory with owner and privileges.
AI agent identity and permissions
Limited public supportRegister AI agents as accountable identities, bind them to owners and delegating users, authorize task- and tool-level access, issue short-lived credentials, review access, and revoke or suspend agent authority.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A test agent is registered with a unique identity, accountable owner, purpose, and permitted resources.
AI coding-agent and workstation security
Limited public supportDiscover and govern AI coding agents, integrated development environment (IDE) assistants, command-line agents, skills, hooks, extensions, Model Context Protocol (MCP) tools, filesystem access, commands, network activity, secrets, and software-supply-chain actions on developer workstations and build environments.
NIST AI RMF PlaybookNIST Cybersecurity Framework 2.0CIS Critical Security ControlsISO/IEC 42001OWASP GenAI Security Solutions Landscape
A test coding agent and its skills, hooks, extensions, or Model Context Protocol (MCP) tools appear in an attributable inventory.
AI cost and usage controls
No supporting claim foundVisibility, attribution, budgeting, rate limiting, anomaly detection, and optimization for AI usage and spend across models, agents, workflows, and owners.
Commercial MetadataNIST AI RMF PlaybookNIST Cybersecurity Framework 2.0ISO/IEC 42001OWASP Agentic AI Security Solutions Landscape
A controlled AI usage event is attributed to user, team, model, workflow, or owner with cost or token metrics.
Licensing model
Limited public supportPublicly discoverable commercial model such as per user, per seat, per app, per token, per integration, or enterprise platform license.
The vendor can map the sourced commercial model to per-user, per-seat, per-app, per-token, per-integration, or platform packaging.
Public sources
Vendor statements and quoted evidence
Showing the first 6 of 19 source records. Open additional records only when needed.
Salt Security materials reviewed did not provide a public claim for workforce discovery of unmanaged AI tools, accounts, users, and usage.
No quoted source text is recorded for this claim.
Salt Security materials reviewed did not provide a public claim for AI-feature discovery across the enterprise business-application environment.
No quoted source text is recorded for this claim.
Salt claims an inventory of AI-agent application programming interfaces (APIs), Model Context Protocol (MCP) servers, and large language model (LLM) integrations with usage and dependency context.
Discover AI agent APIs, MCP servers, and LLM integrations
Salt Security materials reviewed did not provide a public claim for blocking, coaching, redirecting, or containing unapproved AI use.
No quoted source text is recorded for this claim.
Salt claims sensitive-data flow visibility and blocking of data exfiltration through agent-connected application programming interfaces (APIs).
Block data exfiltration & attacks
Salt Security materials reviewed did not provide a public claim for session-level browser or software as a service (SaaS) controls for employee AI use.
No quoted source text is recorded for this claim.
Show 13 additional evidence records
Salt claims real-time protection against abuse of application programming interfaces (APIs) and Model Context Protocol (MCP) interactions used by AI agents.
real-time protection against AI agent abuse
Salt Security materials reviewed did not provide a public claim for AI inventory, risk, approval, exception, and compliance workflows.
No quoted source text is recorded for this claim.
Salt Security materials reviewed did not provide a public claim for AI-specific adversarial testing and release assurance.
No quoted source text is recorded for this claim.
Salt Security materials reviewed did not provide a public claim for model and AI-artifact supply-chain inspection.
No quoted source text is recorded for this claim.
Salt claims default Model Context Protocol (MCP) guardrails that enforce safe AI-agent behavior.
built-in guardrails, enabled by default, enforce safe agent behavior automatically
Salt claims visibility into agent-driven actions and maps them to application programming interfaces (APIs), methods, and workflows.
visibility into every agent-driven action
Salt claims controls that detect and address risky exposure in Model Context Protocol (MCP) and agent-to-agent (A2A) environments.
the riskiest exposures in MCP and A2A environments
Salt Security materials reviewed did not provide a public claim for non-human identity and machine-credential lifecycle controls.
No quoted source text is recorded for this claim.
Salt reports agent access-governance policies that control which agents may call which application programming interfaces (APIs).
control which agents could call which APIs
Salt Code claims real-time policy enforcement while developers generate code with AI coding assistants.
Security policies are applied in real time as developers generate code.
Salt Security materials reviewed did not provide a public claim for AI usage attribution, budgeting, anomaly detection, and cost controls.
No quoted source text is recorded for this claim.
Salt stated that current customers receive Salt Code at no additional cost under their existing license.
Current Salt Security customers receive it at no additional cost
Salt positions its control layer across application programming interfaces (APIs), Model Context Protocol (MCP) integrations, and deployed AI agents.
across APIs, MCP integrations, and agents in production