Connect
Bring the files, messages, and requirements around a project into one useful conversation.
3rd Meridian exists to give engineers their time back. Relay is an AI engineering assistant that connects the context across CAD, email, documents, and the design lifecycle.
Every day, engineers make decisions across a web of files, messages, requirements, and people. Too often, their scarce attention is spent stitching that web together.
We’re building tools that understand the shape of engineering work, so people can spend their mental energy on the problems only they can solve.
Your AI, with the context engineering work needs.
Relay connects to the tools you already use and helps your AI find what changed, understand what it affects, and prepare what needs to happen next. Your connected data stays within the access boundaries you control.
Plug Relay into ChatGPT, Claude, or your own self-hosted AI.
Tell it what you’re working on. Relay learns from the context you already have.
Get a clear change brief, direct impacts, and next steps before anything consequential happens.
Relay is in development. Explore the planned ways to use it.
Marketplace and repository links will appear here at launch.
If you’re exploring enterprise software for engineering context, documentation, and lifecycle intelligence, talk to us. We’ll listen to how your teams work and help define what Relay should become inside your organization.
Each capability will be paired with a short walkthrough so you can see the work before you try it.
Bring the files, messages, and requirements around a project into one useful conversation.
See the relevant change, its direct dependencies, and the evidence behind the brief.
Review what Relay prepared, approve consequential actions, and keep judgment with the engineer.
Scroll through a Relay workflow. The explanation changes with the moment in the work, while the video stage stays with you.
Relay gathers the files, requirements, messages, and project context an engineer already uses.
Describe what you are trying to understand. Relay uses the connected context instead of making you maintain another system.
Relay traces direct dependencies, labels the evidence, and prepares a brief before anything consequential happens.
Review the recommendation, fill the missing evidence, and approve the next action yourself.
Synthetic examples grounded in the Relay playbook. They show the kind of work Relay is being built to support.
A mounting requirement for a motor housing is revised. Relay identifies the revision, traces the affected interface and dependent components, and prepares a review brief.
An engineer is comparing a welded steel bracket with an aluminum casting. Relay frames the decision around loads, fatigue, corrosion, volume, tooling, inspection, and service.
These are synthetic workflow examples, not customer results or claims about a live integration.
These synthetic walkthroughs show how Relay can help an engineer carry context through a component’s full lifecycle.
1. Frame the decision. An engineer needs a quieter, durable final-drive gear for an electric commuter motorcycle. Relay identifies the decision, the current product revision, the duty cycle, packaging envelope, noise target, service life, and safety consequence. It flags missing torque-speed history before treating a catalog value as a project input.
2. Allocate requirements. Relay turns “quiet and durable” into measurable targets: transmitted torque, speed range, ratio, allowable backlash, tooth contact, temperature, corrosion environment, target life, and inspection method. Each target is linked to its source and verification condition.
3. Compare architecture. It compares spur, helical, and alternative ratio arrangements against axial load, efficiency, noise, packaging, lubrication, manufacturability, and service access. The recommendation remains a decision brief; the engineer owns the architecture choice.
4. Design the component. Relay carries the selected ratio, tooth geometry, face width, shaft interface, bearing loads, seals, lubrication path, material and heat-treatment assumptions into the gear definition. It keeps the gear, shaft, bearing, housing, and lubricant interfaces visible.
5. Analyze and verify. It can organize screening calculations for tooth bending, contact stress, shaft torque, thermal load, and bearing reaction when inputs and validity limits are explicit. It separates calculated screening from qualified rating, fatigue life, noise testing, and release evidence.
6. Change and release. If torque or packaging changes, Relay traces the direct dependency path to ratio, tooth geometry, shaft, bearings, housing, lubricant, drawings, test plans, supplier notes, and service instructions. It prepares the change brief, asks for approval, and documents the approved decision after the engineer edits the CAD.
1. Frame the decision. A power-electronics team needs a cooling plate that removes a known heat load while staying inside temperature and pressure-drop limits. Relay separates observed operating data from inferred duty cycle and identifies the missing fluid properties, inlet condition, allowable pressure drop, fouling assumption, and manufacturing route.
2. Build the requirement chain. Heat duty becomes thermal resistance, surface temperature, flow rate, pressure drop, leak-tightness, service life, cleanliness, and inspection requirements. Relay links each requirement to the component, seals, fittings, pump, control limits, and verification method.
3. Explore concepts. It compares machined channels, brazed plates, and formed or bonded passages using heat transfer, pressure drop, material compatibility, joining risk, tooling, volume, inspection access, and repairability. It labels generic correlations as heuristic until geometry and flow regime are confirmed.
4. Design and analyze. Relay carries channel dimensions, wall thickness, manifolds, ports, seals, material, surface treatment, and joining assumptions into the model context. It can organize first-order heat and flow checks, expose the equations and validity regime, and identify where CFD, coupon testing, burst testing, or thermal cycling is needed.
5. Release with evidence. When a supplier proposes a different alloy or brazing process, Relay compares the proposal against corrosion, thermal expansion, pressure boundary, inspection, and qualification evidence. It prepares the open items instead of silently accepting supplier capability as fact.
1. Frame the decision. A robot wrist bracket must carry payload and acceleration loads without losing positioning accuracy. Relay gathers workspace, payload, cycle time, acceleration, joint torque, stiffness, repeatability, environment, cable routing, and safety-state context.
2. Allocate interfaces. It maps the bracket to the actuator flange, bearing, end effector, fasteners, cable path, guarding, and calibration reference. It makes the load path and tolerance chain visible before geometry is optimized.
3. Choose a design route. Relay compares machined aluminum, steel, and topology-informed cast or printed options against stiffness-to-mass, fatigue, vibration, datum stability, production volume, inspection, and repair. It keeps manufacturing assumptions attached to the candidate geometry.
4. Check and test. It organizes static, modal, fatigue, fastener, clearance, and tolerance-stack screening, then connects each result to the required FEA, metrology, cycle test, and safety review. It flags that a first-order result does not qualify a human-consequence design.
5. Learn across the lifecycle. A field vibration issue can be linked back to the bracket revision, joint loads, fastener evidence, inspection record, and maintenance action. Approved lessons can feed the next requirement and design iteration.
1. Receive the signal. A message in Slack, Teams, or email says that a mounting requirement has changed. Relay identifies the exact source message, project, artifact, and revision instead of guessing from a similar filename.
2. Connect context. It retrieves the relevant CAD assembly, requirement document, drawing, test result, and prior decision record. The engineer can ask naturally from ChatGPT, Claude, or a self-hosted client while the source and revision remain visible.
3. Trace impact. Relay follows the direct path from changed requirement to interface, component, drawing, analysis, test, supplier note, and maintenance documentation. It returns the shortest useful path first and keeps unrelated connected knowledge out of the response.
4. Prepare, don’t silently act. Relay drafts a change brief, lists evidence and open questions, proposes documentation and communication updates, and asks for approval before consequential actions. The engineer makes the CAD edit and remains responsible for engineering judgment.
5. Confirm and document. After the revised file is detected, Relay checks that the intended change is present, records the approved decision and evidence, updates the chosen project records, and drafts a clear message for the affected team. It does not publish or release a change without the required human approval.
All examples are synthetic workflow demonstrations based on the current Relay playbook. They are not customer results, certified analyses, or claims that every integration is live.