A two-code mechanical CPC is the patent equivalent of a clean engineering drawing. Boston Dynamics' grant US12103164B2 ("Gripper mechanism," issued October 1, 2024; inventor Brian Todd Dellon) classifies only into B25J 15/0028 and 15/0042 — both gripping-head structure codes. No control, no perception, no learning. The invention is the gripper's mechanics, and the claim says so by where it sits.

But the surprise is in claim 1's preamble: this is not a bare gripper claim, it is a robot claim. Claim 1 recites "a body; four legs coupled to the body," each leg jointed "for articulation of the leg relative to the body," plus "an arm coupled to the body" and, at the arm's distal end, the gripper. In other words, the independent claim fences the gripper as the hand of a quadruped-plus-arm — the Spot-with-manipulator configuration — not as a standalone end-effector. That framing matters for infringement: it reads on the integrated mobile manipulator, which is precisely the product Boston Dynamics sells.

The mechanical heart of the claim is a specific kinematic chain. A "linear actuator" is "coupled to the first jaw member through a pivot body and a cam," and "actuation of the linear actuator rotates the cam to rotate the first jaw member relative to the second jaw member about a first axis." That is the primary grasp motion: linear push converted, through a cam, into jaw rotation. The non-obvious twist is the second degree of freedom: "the pivot body is configured for rotation about a pivot axis to allow for rotation of the first jaw member about the pivot axis." So the same jaw both closes (driven by the actuator through the cam) and pivots (about an independent axis on the pivot body) — claim 11 makes this explicit, reciting that pivot-body rotation "permits rotation of the first jaw member in a direction that is different than" the actuator-driven direction. A two-motion jaw from a single linear actuator is the inventive geometry.

“A gripper mechanism includes a pair of gripper jaws, a linear actuator, and a rocker bogey. The linear actuator drives a first gripper jaw to move relative to a second gripper jaw.”— U.S. Patent No. 12,103,164 source

The dependent claims tighten the mechanism into a buildable assembly. Claim 3 specifies the linear actuator as "a lead screw, a motor…and a ball nut positioned on the lead screw" so the nut translates as the screw turns, with "a protrusion associated with the ball nut…positioned within a slot of the cam" — that protrusion-in-slot is how linear travel drives cam rotation. Claims 4 and 5 shape the slot ("curved," and specifically "an involute slot," an involute profile being the standard way to get smooth, controlled motion transfer). Claim 6 adds "a carrier" with "a carrier slot extending along a linear axis" that constrains the protrusion to translate cleanly. Claims 7–10 then fix the geometry in space: the pivot body pivots on the ball nut (claim 7), the pivot axis is "orthogonal to the axis of the lead screw" (claim 8), the protrusion axis is "orthogonally oriented with respect to the pivot axis" (claim 9), and all three axes — lead screw, pivot, protrusion — are mutually orthogonal (claim 10). That mutually-orthogonal three-axis arrangement is the precise structure the patent fences. (The abstract's "rocker bogey" maps to the pivot-body/cam assembly that lets the jaw both close and pivot.)

In manipulation IP, mechanism claims are the bedrock — you can change the controller or retrain the policy, but you cannot software your way around a claimed mechanical structure. A jaw that gets two motions from one actuator, via an involute cam slot and a triorthogonal axis set, is unusually hard to design around if the geometry is genuinely novel, because the orthogonality and the cam profile are what make the dual motion work. Claims 12–22 re-fence the same structure as an actuation method on the same quadruped-plus-arm robot, so the enforceable surface spans device and method.

For the manipulation beat, this fits Boston Dynamics' shift toward commercial manipulation — the Stretch warehouse robot and the new electric Atlas both need grippers, and a company moving from locomotion demos to manipulation products needs end-effector IP. A clean mechanism grant whose independent claim already nests the gripper in a legged-robot-plus-arm is exactly the foundational hardware asset that posture requires.

From a portfolio angle, read this beside Boston Dynamics' 2025 mobile-manipulator grants (US12251831B2, US12240105B2, US12186919B2): the company is assembling a manipulation portfolio spanning the gripper, the arm, the mass-estimation control, and the perception mast. The gripper-mechanism patent is the hand in that system; together they fence an integrated mobile-manipulation product, not just a part.

Caveats. Gripper mechanisms have deep prior art; the grant turns entirely on the specific structure recited in claim 1 and tightened by claims 3–10 — the cam-slot drive and the mutually-orthogonal lead-screw/pivot/protrusion axes. The two-code classification tells you it is mechanical; the dependent claims tell you the geometry. Note too that claim 1 recites a four-legged robot, so a bare bench gripper may not read on it — that integration is both reach and limit of the fence.

For the file: a foundational gripper-mechanism grant whose independent claim already sits on a quadruped-plus-arm, inside Boston Dynamics' emerging mobile-manipulation portfolio. Pull US12103164B2 alongside the 2025 manipulator family in the patent record, read claims 1, 3, and 9–10 for the dual-motion jaw and orthogonal-axis structure, and treat the cluster as an integrated-product fence.