Robots Center Agents Network
Log in Create workspace

Robots Center · Operator console

The control plane for AI agents and robot fleets

AI agent operations and robot fleet management in one workspace. Monitor your agent fleet and robot fleet through reported traces, heartbeats, and telemetry. Control credentials and permissions for AI agents in software workflows and on robots.

Create workspace API Docs

Demo topology

7 nodes · 7 routes

Fleet cohorts

Zero-touch claim

Emergency freeze

Telemetry rollups

Five surfaces

Before, during, and after every run

Explore AI agent operations solutions →

01 · Identity and policy

Agent Auth & Permissions

Scoped identities, managed secrets, and installable policy packs stand between your agents and external tools, so risky actions are explicitly controlled.

Explore Agent Auth & Permissions →
  • One primary workspace API key for direct public API access.
  • Connector grants with allow and deny effects.
  • Approval policies with reviewer routing and expiration windows.
  • Break-glass freeze on a workspace, a connector, or a single agent, with a required reason and optional auto-thaw.
  • Audit events for auth checks, approvals, freezes, and policy decisions.

02 · Runtime visibility

Observability & Replay

Runtime, eval, and replay traces land in one data model with failure clustering, completion-time anomaly windows, and related-record correlation.

Explore Observability & Replay →
  • Cross-surface search plus exact-match drilldowns for environment, target, agent, request ID, and external trace ID.
  • Completion-time anomaly alerts with full grouped inbox visibility, delivery history, and operator lifecycle actions.
  • Workspace-configured read-time payload redaction for operator views.
  • Replay jobs with variant labels, config overrides, and semantic diff output.

03 · Release quality

Evals & Version Review

Workflow targets, versions, suites, and runs live in the same app, so operators compare evaluation evidence and record promotion decisions before deploying through their own release process.

Explore Evals & Version Review →
  • Rubric-backed eval suites and scheduled runs.
  • Shadow tests and baseline regression comparisons for candidate versions.
  • Review queues for failed and LLM-judged eval cases.
  • Incident-generated eval drafts that can be promoted into real suites.

04 · Coordination

Cross-Agent Communication

Agents message each other natively through the platform, with task delegation, group coordination, and discovery layered on one realtime protocol.

Explore Cross-Agent Communication →
  • Durable direct messages with acknowledgement and redelivery for offline peers, subject to expiry.
  • Task delegation with priority levels and automatic retries.
  • Agent groups with broadcast coordination for multi-agent workflows.
  • Deterministic agent discovery by exact advertised capability string.
  • Realtime WebSocket channel for agents at /socket.

05 · Fleet operations

Agents & Robot Fleets

Software agents and physical robots sit in one registry with shared cohorts, status, and alert semantics, so operators filter one board by group, site, or tag.

Explore Agents & Robot Fleets →
  • Cohorts for groups, sites, and free-form tags, with static membership or rules that match on model, firmware, capabilities, and tags.
  • Zero-touch provisioning: CSV pre-registration, multi-claim enrollment tokens with serial allowlists, and per-claim unique credentials.
  • Batched telemetry ingest with minute and hour rollups behind 1h, 24h, 7d, and 30d trend charts per robot and per cohort.
  • Audited bulk operations across a filtered selection, with live progress, per-target retry, and cancel.

Integration surface

The app exposes control-plane APIs around your runtime.

Agents send

  • Use the primary workspace API key directly on protected public APIs.
  • Ask whether a tool action is allowed, denied, or requires approval.
  • Send trace events for runtime, eval, and replay execution.
  • Queue eval runs, shadow tests, and replay variants.
  • Claim a fleet enrollment token on first boot and receive a credential of their own.
  • Push batched, idempotent telemetry readings that survive an offline backlog.
  • Send and receive agent-to-agent messages, delegate tasks, and coordinate groups over /socket.

Operators get

  • Approval queues with explicit reviewer actions and audit trails.
  • Trace timelines, global search, grouped failures, and replay comparisons.
  • Eval run status, pass rates, review queues, and case-level results.
  • One live board for agents and robots, filtered by cohort, site, or tag, patched in place as events land.
  • Telemetry trends, bulk operations with per-target results, and a break-glass freeze.
  • Workspace controls for memberships, data policies, webhooks, and identity.

Own your agent. Monitor your fleet.

Give each agent an identity, decide what it can access, and follow its reported activity. Start a workspace with no credit card required.

Create workspace API Docs

Already have a workspace? Sign in