CydraLabsZero trust for AI agents

Agent security · Zero trust · Continuous governance

Control every consequential action your AI agents take

Discover agents, map their effective authority, enforce policy before tools execute, and produce evidence for every decision across cloud, model and framework environments.

Request a demo Assess your agent risk

Built for enterprises that need AI agents to act without giving up security, accountability or control.

Action decision

Illustrative
Agent
agent://acme/sales-agent
Delegated user
j.smith (Sales)
Tool
crm.export_contacts
Target
Production · Confidential
Agent Risk Score
59 · High
Inspection
Personal data detected

Decision

Approval required before execution

Decision recorded in the hash-linked evidence chain

Export not executed: waiting for a named approver

Illustrative example of a CydraGateway decision. A sales agent acting for j.smith requests a CRM contact export from a confidential production target. Personal data is detected, so approval is required before execution. The decision is recorded as evidence; the export has not run.

The problem

Enterprise agents are gaining authority faster than security can see it

Each agent can reach tools and data through identities that security teams did not design for it.

Unknown agents

Agents are built on many clouds, models and frameworks, often without reaching an inventory or a security review.

Excessive permissions

Delegated user scopes, shared secrets and every reachable tool give agents far more authority than their task needs.

Fragmented controls

Each framework and model platform implements guardrails differently, if at all, and none of them share policy.

Ungoverned tools

MCP servers, plugins and APIs are added by individual teams, and tool calls bypass controls built for humans.

Weak evidence

When an auditor or responder asks why an agent was allowed to act, logs rarely show the policy, context and approver.

Platform

One continuous control loop

CydraShield, CydraGateway and CydraGovern share the same identity, policy, approval, telemetry and evidence services, working as one loop.

Explore the platform

  1. Inventory agents, models, MCP servers, tools and the identities they use.

  2. Evaluate each proposed action against policy, risk and context before it runs.

  3. Allow, deny or require approval, then execute with scoped, short-lived credentials.

  4. Monitor actions, behaviour and changes in effective authority over time.

  5. Record every decision as signed, hash-linked evidence mapped to controls.

  6. After Prove, the loop returns to Discover.

After Prove, evidence and findings feed the next round of Discover: the loop is continuous.

Core products

CydraShield, CydraGateway and CydraGovern

One clear outcome each. Every capability carries a status label: Available, Beta, Preview or Roadmap.

Security platform for AI agents

CydraShield

Discover agents, map authority and contain risk

CydraShield capabilities

  • Agent inventory and ownership

    Available

  • MCP server and tool discovery

    Beta

  • Model inventory (connector data)

    Available

  • Live model discovery from provider APIs

    Preview

  • Prompt and activity logging

    Available
Explore CydraShield

Secure agent, MCP and tool gateway

CydraGateway

Authorise every tool call before execution

CydraGateway capabilities

  • Identity-aware agent gateway

    Available

  • MCP, API and tool mediation

    Available

  • Pre-execution policy decisions

    Available

  • Human approval gates

    Available

  • Data and secrets inspection

    Available
Explore CydraGateway

AI governance and compliance

CydraGovern

Turn controls into continuous evidence

CydraGovern capabilities

  • AI-system classification

    Beta

  • Control mapping across frameworks

    Beta

  • PCI DSS control coverage

    Roadmap

  • Evidence linked to runtime decisions

    Available

  • Exceptions and remediation tasks

    Available
Explore CydraGovern

Extensions

Extensions build on the core products and are on the roadmap. About the extensions

Action control

Every consequential action is decided before the tool executes

CydraGateway puts a deterministic decision point between agents and enterprise tools, the same for every framework and model provider.

  1. Stage 1

    Identity and context

    Authenticate the agent, resolve the delegated user, tool and target, and attach data, risk and graph context.

  2. Stage 2

    Policy and risk

    Evaluate versioned, deterministic policy and inspect content. Errors fail closed.

  3. Stage 3

    Approval

    A named approver issues a single-use token bound to the exact request.

  4. Stage 4

    Scoped execution

    The action runs once, with a per-execution credential limited to the target.

  5. Stage 5

    Evidence

    The decision and result are appended to a signed, hash-linked evidence chain.

Illustration, not a live request

Pauses for a human decision. Execution waits for the illustrative approval, then evidence is recorded.

  1. 1Receive PassedTool call submitted with an idempotency key.
  2. 2Authenticate PassedWorkload token valid for this tenant and audience.
  3. 3Resolve PassedOwner, delegated user, tool and target resolved.
  4. 4Context PassedData classification, risk score and graph context attached.
  5. 5Policy PassedAn approval rule matches for confidential data.
  6. 6Inspect HeldPersonal data detected in the requested export.
  7. 7Decide HeldDecision: require approval. Nothing runs yet.
  8. 8Approve ApprovedIllustrative approval: a named approver signs a single-use token bound to this request.
  9. 9Execute ExecutedRuns once, after approval, with a scoped credential.
  10. 10Evidence RecordedDecision, approval and result appended to the evidence chain.
  • Policy errors fail closed
  • Deny takes precedence over approval and allow
  • Every outcome produces an evidence record
All ten steps explained

Security Graph

See effective authority, not just permissions

The graph connects users, agents, models, MCP servers, tools, data and actions.

Effective authority
What each agent can reach directly and transitively, through delegated users and tools.
Blast radius
Everything exposed if an agent or MCP server is compromised.
Prioritised remediation
High-risk paths ranked with the Agent Risk Score, so you fix what matters most.

Status: Available

delegatesusesconnectscan callexposesreadsreachesUserDelegating humanSales agentagent://acme/salesModelFoundation modelMCP serverUnreviewedCRM toolwriteCustomer dataConfidentialEmail sendExternal egressHigh-risk path (dashed)

High-risk path

User → sales agent → unreviewed MCP server → customer data and external email: a possible exfiltration path.

Select a node to see its relationships.

Graph as text
  • User (Delegating human): delegates Sales agent (high risk).
  • Sales agent (agent://acme/sales): uses Model; connects MCP server (high risk); can call CRM tool; User delegates it (high risk).
  • Model (Foundation model): Sales agent uses it.
  • MCP server (Unreviewed): exposes Email send (high risk); reaches Customer data (high risk); Sales agent connects it (high risk).
  • CRM tool (write): reads Customer data; Sales agent can call it.
  • Customer data (Confidential): CRM tool reads it; MCP server reaches it (high risk).
  • Email send (External egress): MCP server exposes it (high risk).
Illustrative example, not customer data.

Solutions

Start from the outcome you need

Solutions for regulated industries

Integrations

Coverage across the systems your agents already use

Integrations are named in text with their status. We use vendor logos after support is verified.

  • Agent frameworks and protocols

    Beta

    Framework-agnostic HTTP API and Python agent SDK; MCP server and tool discovery (Beta).

    Planned: Named framework adapters and A2A protocol controls.

  • Model providers

    Preview

    OpenAI and Anthropic discovery connectors (mock adapters Available; live APIs Preview).

    Planned: Google (Gemini), open-weight and customer-hosted models through the model router.

  • Cloud platforms

    Available

    Reference deployment on AWS (EKS, RDS, ElastiCache, S3) and Docker Compose.

    Planned: Agent and AI-service discovery in AWS, Azure and Google Cloud; AWS Security Hub.

  • Identity

    Preview

    Microsoft Entra ID discovery connector and OIDC sign-in; Entra ID and Okta sign-in in Preview.

    Planned: Okta discovery connector.

  • Security operations

    Beta

    Signed webhooks (Beta) and OpenTelemetry export of traces.

    Planned: Microsoft Sentinel, Splunk, Google Security Operations, CrowdStrike and Microsoft Defender.

  • IT service management

    Roadmap

    Not yet available.

    Planned: ServiceNow and Jira for approvals, findings and remediation tasks.

  • Enterprise systems

    Preview

    GitHub discovery connector (mock adapter Available; live API Preview).

    Planned: Salesforce, SAP, Microsoft 365 and GitLab.

Product names identify interoperability targets and do not imply partnership or endorsement.

Deployment and trust

Deploy where your data must stay, and verify how it is protected

Statements describe the current platform design. They are not certifications or third-party attestations.

Deployment

  • CydraLabs-managed SaaS

    Hosted and operated by CydraLabs.

    Pilot programme

  • Customer cloud or private VPC

    Deployed into your own cloud account with the CydraLabs reference deployment.

    Pilot programme

  • On-premises or sovereign deployment

    Planned for environments that cannot use cloud services.

    Roadmap

Trust

  • Privacy controls

    Payloads are stored as hashes plus redacted previews by default; logs redact secrets and personal data.

  • Encryption

    TLS in transit. In the AWS reference deployment, database, cache and object storage are encrypted at rest with KMS keys.

  • Assurance

    Tenant isolation with row-level security, fail-closed enforcement and signed, hash-linked evidence. No third-party certifications yet.

Visit the Trust Centre
Design-partner programme

Customer evidence

Pilot outcomes, case studies and quantified benchmarks will appear here when they are available and design partners agree to publish them. Until then this section describes the design-partner programme; it is not customer evidence, and we do not display customer logos, quotes or results.

Apply to become a design partner

What design partners receive

  • Early access to the proof-of-concept platform
  • Direct input into roadmap priorities
  • Joint threat modelling of your agent estate
  • Working sessions with the engineering team

What we ask

  • Real agent use cases, in a non-production environment
  • Regular feedback sessions
  • A named security or platform sponsor
  • Candid assessment of what does not work

Benchmarks

Gateway latency and detection benchmarks are not yet published. We will publish the methodology with the results.

Pricing and packaging

Choose the right level of protection for your AI-agent environment.

The prices below are indicative prices for the core CydraLabs Agent Trust and Action Control Platform. They are not prices for each individual CydraLabs product and do not provide unrestricted access to the entire portfolio.

  • Developer

    Pilot programme

    Free

    Managed agents
    Up to 5 non-production managed agents
    Best for
    Design-partner engineers, evaluators and Academy learners
    Start free

  • Team

    Pilot programme

    £499per month

    Managed agents
    Up to 50 managed agents
    Best for
    Small teams and early pilots
    Start a pilot

  • Business

    Pilot programme

    From £2,500per month

    Managed agents
    Up to 500 managed agents
    Best for
    Production business deployments

    Support options confirmed during quotation

    Request a demoRequest a quote

  • Enterprise

    Pilot programme

    £50,000 to £300,000+ARR

    Managed agents
    Negotiated
    Best for
    Multi-team or regulated enterprises

    Support options confirmed during quotation

    Request a tailored quote

  • Product add-ons

    Roadmap

    Priced separately

    Managed agents
    Scoped per add-on
    Best for
    Customers extending the core platform
    Discuss product add-ons

Prices apply to the core CydraLabs platform. Premium connectors, private deployment, extended evidence retention, professional services, specialist product modules and usage above the included limits are priced separately. Final limits and prices remain subject to pilot validation, deployment requirements and contract terms.

See full pricing and packaging

Resources and Academy

Learn, evaluate and go deeper

Put a control point in front of your agents' actions

See CydraShield, CydraGateway and CydraGovern working together on a realistic agent estate, or start with an indicative risk self-assessment.

Request a demoAssess your agent risk