Binary³ Agentic Security · Agent Capability Layer

Give your AI agents
security tools.

Binary³ MCP is a developer access layer that exposes Binary³ security capabilities to MCP-capable AI clients through one machine endpoint — any client that supports remote MCP servers over HTTP with a standard bearer Authorization header, including custom agent systems. Keep the agent in its workflow while KeySweep and MicroDefend return structured, redacted results.

Explore the tools Read the docs

MCP ENDPOINT — machine-facing protocol surface, not a webpage
https://mcp.binarycubed.com/mcp

Client compatibility: standard MCP over HTTP with an Authorization: Bearer header — no custom-header requirement. Claude, GPT, Cursor, Gemini-class clients are protocol-compatible; per-client behavior is not yet independently qualified.

MCP-capable client / custom agent
Binary³ MCP
Binary³ Security Capabilities
KeySweep · MicroDefend

MCP routes requests to the security capabilities. Axon — a separate Binary³ agent-security capability — is not in this execution path.

What MCP gives your agent

Real security operations,
as native tool calls.

One connection

A single MCP endpoint configuration makes the Binary³ tool surface discoverable to your client through standard protocol discovery.

Structured results

Tools return typed, redacted output — detection types, severities, and findings — never raw secret values.

Scoped access

Each credential carries exact per-tool entitlements. A credential without a tool's entitlement cannot invoke it.

Supported Binary³ tools

Two focused operations.
No invented surface area.

KeySweep

keysweep_scan_text(text: string)

Inspect supplied text for exposed secrets — API keys, tokens, private keys, credentials, database URLs. Pure analysis with no network access; detected secret values are redacted from the output.

Tool reference →

MicroDefend

microdefend_scan_website(url: string)

Scan an authorized public website for TLS posture, security headers, metadata, and exposed paths. Returns status, scan_id, findings, and a redacted result.

Tool reference →

How it works

Connect. Discover. Invoke.
Analyze. Return.

Connect

Configure a compatible MCP client with the Binary³ endpoint and your issued credential.

Discover

The client discovers the security tools Binary³ makes available — the catalog always reflects the registered tool surface.

Invoke

The agent invokes a tool with its typed input. Authentication, entitlement, and tenant binding are verified before any work runs.

Analyze

The underlying Binary³ service — KeySweep or MicroDefend — performs the actual security operation.

Return

Structured, redacted results return to the calling agent or client.

Architecture

An access layer —
nothing more claimed.

Clients

MCP client
HTTP transport
Custom agent

Access layer

Binary³ MCP
mcp.binarycubed.com/mcp

Security capabilities

KeySweep
MicroDefend

Axon is a separate Binary³ agent-security capability — explore Axon.

MCP is a developer access and integration layer: it exposes security capabilities to agents. It is not itself a scanner, a risk engine, or an enforcement engine — the security value comes from the Binary³ capabilities it exposes.

Qualification status: the public JWKS surface is production-qualified. Authenticated MCP end-to-end operation is independently qualified (post-roadmap production qualification gate, 2026-09-14).

Full security model →

Developer quick start

Connect in five steps.

Choose your MCP client

Any client that supports remote MCP servers over HTTP and can send a standard bearer Authorization header, or a custom agent runtime. Named-client compatibility is not yet independently qualified; the configuration shape below is the protocol-standard form.

Point it at the Binary³ endpoint

The machine-facing protocol endpoint:

https://mcp.binarycubed.com/mcp

Authenticate

Send your issued bearer credential on every request. Never commit it or place it in browser code.

Authorization: Bearer <BINARY3_MCP_API_KEY>

Discover tools

Your client lists the available tools and their input schemas from the endpoint.

Call a security capability

Invoke keysweep_scan_text or microdefend_scan_website and read the structured, redacted result.

See a simulated invocation →

Access

Commercial terms,
stated plainly.

MCP plan pricing is not yet published, and live billing is not active. We will not manufacture tiers or quotas before they exist. Access conversations start with your intended tools and agent environment.

Access status
Documentation

The practical docs.

Connection configuration, authentication, tool contracts, safe result fields, error states, and the public JWKS surface.

Read the docs

One endpoint.
Real security tools.

See RQE plans Browse the tools