Agility Robotics Digit (warehouse humanoid / MMR)
Digit is positioned less as a “general household humanoid” and more as an industrial product: a human-scale robot built to walk in standard warehouse aisles, pick up and move totes, and integrate into existing workflows without forcing a facility redesign. The core story is systems closure—hardware + safety + fleet operations + commercial deployment—where repeatability matters more than viral athletic demos.
The easiest way to misread Digit is to judge it as “a humanoid that isn’t trying to do everything.” That’s actually the design choice. Digit’s bet is that the first real humanoid business is not a kitchen assistant—it’s warehouses and distribution centers, where labor is expensive, turnover is constant, and the work is physically repetitive. In that environment, “good enough” manipulation paired with robust mobility can create immediate value if the system can run day after day with predictable safety and predictable cost.
Digit is typically described as a bipedal mobile manipulation robot (MMR). In practice, that means two legs for navigating human-centric layouts (aisles, thresholds, tight spaces) and arms/end effectors aimed at a small set of high-frequency jobs like tote transfers and tote recycling workflows. The form factor matters because warehouses are built around people. A solution that fits the space reduces integration friction, which is often the real killer of automation projects.
What makes Digit “advanced” isn’t only the body—it’s the operational layer. Agility markets Agility Arc as a cloud automation platform to deploy and manage Digit fleets, which signals an important product posture: Digit is treated as a managed system, not a one-off robot. If you can onboard a site, define tasks, monitor health, push software updates safely, and measure throughput reliably, you have something that resembles an industrial product rather than a lab demonstration.
The commercial signals are unusually concrete for a humanoid: a multi-year RaaS agreement with GXO, a public “first commercial operations” framing for June 5, 2024 at a GXO facility, and later commercial expansion stories (including Mercado Libre) that emphasize real work volume rather than hype. This is the kind of evidence investors and operators care about: not “can it do it,” but “can it do it repeatedly with a safety case and a deployment playbook.”
The honest constraint: Agility shares fewer deep component-level disclosures than some rivals, and third-party spec lists can drift. For a rigorous robotics analysis, the correct move is to separate (1) publicly confirmed job envelope and deployment milestones from (2) speculative performance claims. Digit is strong where it matters most right now—systems closure in a specific domain—even if the broader “general humanoid” story is still unfolding.
Scoreboard
82 / 100
66 / 100
86 / 100
84 / 100
83 / 100
72 / 100
Technical specifications
| Spec | Details |
|---|---|
| Robot owner | Agility Robotics |
| Category | Bipedal mobile manipulation robot (MMR) for warehouses and distribution |
| Standing height | 5' 9" (public “Digit by the Numbers”) |
| Payload | 35 lbs (public “Digit by the Numbers”) |
| Operational range | 5.5 ft (public “Digit by the Numbers”) |
| Autonomous docking & charging | Yes (public “Digit by the Numbers”) |
| End effectors | Customizable end effectors are highlighted publicly; the product posture is workflow-driven tooling rather than “one hand for all tasks.” |
| Fleet platform | Agility Arc (cloud automation platform for deploying and managing Digit fleets) |
| Commercial model | Robots-as-a-Service (RaaS) is central to early deployments and sets accountability around uptime and throughput. |
| Safety milestone | Digit received an OSHA-recognized NRTL-related safety milestone enabling broader real-world use cases (as described by Agility). |
| Speed / runtime / full DoF disclosure | Not consistently published in primary sources. Treat third-party spec sheets as tentative unless backed by Agility documentation. |
| Primary deployment path | Fulfillment centers, putwalls, tote recycling workflows, and “automation seam” transfers (AMR ↔ conveyor ↔ station). |
| Public spec reliability | High for job envelope + height/payload/range + commercial milestones; medium for deep component/performance numbers. |
What stands out beyond the scoreboard
- Domain focus: Digit is not trying to win every humanoid benchmark. It’s tuned for logistics work where ROI is measurable and workflows are repeatable.
- System closure: Agility Arc positions Digit as a deployable fleet product (onboarding, monitoring, operations), not a custom-engineering project per site.
- Commercial proof: A multi-year RaaS agreement and a public “commercial operations” milestone reduce the gap between prototype and workforce.
- Safety pathway: OSHA-recognized NRTL-related safety milestones signal real hazard analysis and field-readiness, which is essential for insurance and scaling.
- Warehouse-native form factor: Human-scale height and reach matter because the environment is already built; the robot fits the facility rather than forcing redesign.
- Throughput economics: The question isn’t “can it move a tote?” It’s cost-per-tote including downtime, resets, and supervision.
- Recovery behavior: Real floors and real congestion cause non-ideal situations. Recovery (without human intervention) is where deployments either scale or stall.
- General dexterity: Tote workflows are a great wedge, but expanding to higher-mix picking and more varied object handling is the long frontier.
- Site variability: “Same warehouse type” still means endless variation: aisle widths, clutter patterns, floor wear, lighting, and human traffic.
- Fleet ops discipline: Scaling requires staged rollouts, telemetry, remote debugging, spare parts logistics, and safe update pipelines.
Provider pick
| Priority | Pick | Why | Tradeoff to accept |
|---|---|---|---|
| Best first environment | Fulfillment centers with clear lanes and defined “handoff” stations | Digit’s strength is seam work—bridging AMRs, conveyors, and stations with a human-scale robot that doesn’t demand facility redesign. | Early success will be scaffolded (marked lanes, defined stations). That’s normal and it’s how you get ROI early. |
| Best first workload | Tote transfer + tote recycling workflows | High-frequency, measurable tasks. You can instrument success rate, cycle time, and intervention rate quickly. | Not glamorous. But boring is exactly what scales and pays. |
| Fastest scaling loop | Repeatable jobs with controlled variability | Controlled variability yields fast learning/iteration without exploding the safety case. It’s ideal for expanding a skill library. | This is not “general intelligence.” It’s disciplined product expansion. |
| Long-term bet | Multi-skill warehouse humanoid under fleet management | If Arc plus deployment playbooks keep maturing, Digit can broaden from totes to more station types without breaking ops. | The last 10% is brutal: reliability, serviceability, and cost-per-hour decide winners more than headlines. |
Real workloads cost table
Digit is best evaluated on “warehouse reality” where the goal is simple: keep material moving with predictable safety and predictable economics. The table below frames cost the way operators feel it: not just the robot price, but the downtime + intervention + exception-handling budget needed to sustain throughput.
| Scenario | Input | Output | What it represents | Estimated cost driver |
|---|---|---|---|---|
| Tote transfer (AMR ↔ conveyor) | Standard totes, known pickup zones | Totes moved reliably between systems | Classic “seam” task—connects islands of automation | Interventions per hour + cycle time variability |
| Putwall support | Defined stations, moderate traffic | Material staged and moved on schedule | Stable, repeatable motion near humans | Safety constraints vs throughput tradeoff |
| Tote recycling workflow | Return totes, repeat paths | Totes routed into the right flow | “Boring work” that proves ROI | Downtime + remote debugging speed |
| Mixed-traffic aisle operations | Humans, carts, dynamic obstacles | Safe navigation + task completion | Recovery behavior under real congestion | Recovery time + near-miss tolerance |
| High-mix manipulation expansion | More object types, more station types | Broader skill library | Where humanoids become truly flexible | Training time + verification + safety re-validation |
How to control cost (a practical playbook)
1) Treat recovery metrics as the real KPI, not demo success
The difference between a “cool robot” and a workforce robot is recovery: what happens when something goes wrong. Track retries, time-to-recover, and escalation frequency as first-class metrics. If recovery is weak, your cost-per-tote explodes.
- Measure mean time between interventions (MTBI) and the distribution (tail events matter most).
- Instrument every intervention cause: perception uncertainty, grasp failures, navigation dead-ends, safety stops.
- Gate expansion of scope until recovery behavior stabilizes across shifts (not just a controlled pilot).
2) Engineer the workflow so the robot can be boring
The fastest ROI comes when the environment is designed for predictable handoffs. Mark pickup zones, standardize tote types, and keep station geometry consistent. This reduces ambiguity and increases throughput without risky autonomy leaps.
- Use consistent tote placement and consistent station heights to reduce grasp variance.
- Define robot lanes and human lanes; reduce mixed-traffic chaos during early rollouts.
- Instrument the site so you can correlate performance with floor conditions, traffic, and shift patterns.
3) Make safety authority explicit and independent
Industrial deployments live or die on safety. Build an architecture where safety limits are strict and separate from “smart” behaviors. A safe slow robot can be improved. An unsafe fast robot gets shut down.
- Conservative speed/force limits near humans; clear stop behaviors and safe standoff distances.
- Auditable logging for events and near-misses; safety reviews that operators can trust.
- Site-specific hazard analysis as a repeatable playbook—not a one-off negotiation.
4) Run deployment like a fleet ops business
Scaling humanoids isn’t only robotics—it’s operations. You need monitoring, staged updates, spare parts, and rapid remote diagnosis. Fleet management (like Arc) is the backbone that makes scale possible.
- Remote health telemetry: actuator temps, sensor status, charging success, drift detection.
- Staged rollouts: canary robots → small cohort → fleet; fast rollback if anomalies appear.
- Spare parts logistics and service SLAs: downtime is your largest hidden cost.
FAQ
Why is Digit focused on totes instead of “general tasks”?
Because totes are a high-frequency unit of work in modern fulfillment. If a robot can reliably move totes between systems, it immediately reduces labor pressure and increases throughput. It also creates a stable platform for expanding skills later. In early humanoids, tight scope is a feature: it makes ROI and safety validation achievable.
What’s the real differentiator: legs, arms, or the software platform?
The differentiator is system closure: a robot that can be deployed, monitored, updated, and scaled safely as a fleet. Legs and arms enable the task, but the platform and ops discipline make it a product.
Why does an OSHA-recognized NRTL milestone matter?
It signals that the robot is being evaluated against real-world safety expectations in real deployments. That matters for insurance, compliance, customer confidence, and repeatability of rollout. It also forces the vendor to treat safety as a core architecture requirement rather than an afterthought.
What would most increase confidence for readers?
Audited operational metrics: mean time between interventions, mean time to recover, cycle time distributions, hours run in commercial deployments, and cost-per-tote including downtime and supervision. The humanoid winners will publish “boring numbers” because those numbers decide scale.
Sources used for this page (primary first)
- Agility Robotics: Our Solution (Digit “by the numbers”: 5'9", 35 lbs payload, 5.5 ft operational range, docking/charging)
- Agility Robotics: GXO signs industry-first multi-year agreement (RaaS + Arc)
- Agility Robotics: Digit deployed at GXO (commercial operations milestone: June 5, 2024)
- Agility Robotics: Beyond the Hype (OSHA-recognized NRTL safety milestone + standards list)
- Agility Robotics: Mercado Libre commercial agreement (San Antonio deployment)
- GXO: multi-year agreement announcement (independent confirmation)
- Digital Commerce 360: Mercado Libre deployment coverage (context + “100,000 totes” claim attribution)