The whole platform

Govern agents, MCP servers, LLM gateways and channels from one portal.

Change a subscription, credential, route or freeze centrally — agents pick it up automatically, no redeploys.

Agents can’t self-discover or self-connect; every link is an explicit grant.

Interactive topology of every agent, MCP, gateway, channel and user, health-sorted.

Wire up agent↔agent, agent↔MCP, agent↔gateway and channel↔agent relationships in the UI.

Click any connection in the map to test the policy between source and destination.

Relay card URL, health checks, visibility and auth method.

Healthy / unreachable / unhealthy, plus active / inactive / frozen states.

Instantly pause a single agent’s traffic with a reason.

Freeze all relay traffic — every agent, caller and webhook — during an incident.

Regenerate OAuth secrets and keys without changing the client ID.

See every capability an agent advertises — and govern it.

Live kanban of agent tasks by lifecycle state, with full event history.

Uptime, request volume, latency, error rate and activity charts.

A chronological trail of every state change per agent.

Discover and one-click-connect verified agents.

Register, version, freeze/unfreeze, rotate credentials, full lifecycle log.

None, bearer, API key, query param or OAuth2 — including user-delegated auth-code flow.

A curated official catalog plus a community marketplace, with uptime and latency metrics.

Copy-paste config for Claude Code, Claude Desktop, Cursor and more.

Aggregate multiple providers behind one gateway.

Route by default/fallback or by agent metadata tag, priority-ordered.

OpenAI, Anthropic, Mistral and more, with credential rotation.

OAuth-backed hubs connecting external systems to your agents.

Enable the Teams bot, download the app manifest, auto-linked channel.

Drag-and-drop enforcement rules — no code required.

Switch to code; parse Rego to visual and back.

Block, mask, warn, log or require human review — on every exchange.

Presidio (23+ entity types) and AWS Comprehend (21 types) with confidence thresholds.

Pattern-match data, throttle per window, and control which tools an agent can call.

Police content you can describe but can’t pattern-match — “full names”, “celebrity names”, “dog breeds” — judged in-line by OpenAI’s Decisions API on every exchange.

Act when content mentions the listed items, or when it’s missing one it must contain, such as a disclaimer. Set how confident the judgement must be, and see every question asked.

If a decision can’t be made the exchange is denied. Every judgement is audited with per-item scores — the content itself is never stored.

Apply rules to specific request and response legs.

Bind policies at tenant, agent, user, channel, MCP or gateway level.

See which broader-scope policies apply before local ones; reorder priority.

Owners, validity windows, hash-tracked versions and full change history.

Real-time rule validation and config preview.

Every intercepted request that triggered a policy.

A central list of flagged exchanges awaiting review.

Source user, sink agent, policy, matched content and the linked trace.

Every approve or reject is auditable — who, when, why.

Track escalations raised, resolved and pending, plus SLA compliance.

Who’s reviewing, and every action taken outside policy.

Every interaction and system exchange captured with rich context.

Flame-graph timelines, spans, durations and correlation IDs.

P1–P5 priority with a full alert lifecycle — ack, resolve, dismiss, reopen.

Silence monitors indefinitely or until a set time, with actor tracking.

Across logs, traces and alerts.

Related events linked by correlation ID.

Inactive agents or users suddenly coming back to life.

An agent calling a tool it has never used before.

Agents binding to new MCP servers.

Flagged by the masking middleware.

Agents with no enforced policy, shadow-mode policies and monitor coverage.

Signed, chained checkpoints across the audit trail.

Optional RFC 3161 anchoring of each checkpoint.

Verify all checkpoints, or a chosen time range, against the hash chain.

Audit events and policy actions as CSV for a chosen window, with a hash-chain verification report.

An obligation matrix built from the EU AI Act, as amended by the Digital Omnibus, applied to each system by tier, role and use case.

CIO, CISO, Head of Risk & Compliance, General Counsel, Head of Technology, AI/Transformation Lead and Business Domain Owner.

Share of traffic blocked, warned or held; logged vs masked vs blocked.

Top agents, users, MCP servers, channels and reviewers.

WAU/MAU stickiness, new vs returning, time-to-first-value.

Error-budget burn, p95/p99 latency, SLOs, fleet and dependency health.

Estate overview, by-channel breakdown, use-case heatmap and top failing flows.

Tokens and spend across every LLM gateway call and Decision API check, hour by hour or day by day.

Break usage down by model, provider, gateway or consumer — agent, MCP server or user — then drill in by a second dimension.

OpenAI, Anthropic and Google list prices out of the box; override any model, or price one that isn’t listed.

Cost is worked out when you look, so a price change re-prices past usage too — with over a year of history kept.

Microsoft Entra ID, Google Workspace and Okta.

Auto-provision users and map IdP groups to swarmd groups.

Granular permissions, membership and versioned config.

Invite, assign roles, deactivate/reactivate, activity history.

Secrets, keys and tokens encrypted at rest and in transit.

Chat directly with any agent or gateway to test.