Rapta vs Sindri (2026)

Rapta reads and controls named devices on the line.

Its Supercoach API covers SuperPod robotic inspection stations (connect, home, move to position, queue and cancel motion sequences, emergency stop), connected torque wrenches (update torque configuration, cancel the current tightening operation), and Basler cameras and autofocus, and RFID and Zebra barcode scanners can be installed.

Output signals go out on GPIO pins to the line's automation. MES and ERP connection runs through the API. Customers build MES and ERP connectors on the API themselves, not as native connectors, and Rapta does not describe credential scoping or custom tools.

Sourcerapta.ai/api-documentation (Supercoach API 1.16.22), rapta.ai/knowledge-base/digital-i-o-signaling-and-connections, rapta.ai/ai-platform/enterprise-integrations; readread 2026-09-15

Secure execution connectors are one of Forge's platform pillars, and Forge executes approved actions safely into legacy systems. Procurement orchestration proposes corrective actions, and the execute step closes tasks end to end. The loop runs to closure inside the customer's tenant. Sindri does not explain how the connectors work, and it names no integration catalog, custom tool support or authenticated action framework.

SourceSindri AI, sindri.ai/en/why-sindri/ai-technology and /solutionsread 2026-09-08

Rapta programs each assembly as a sequence of steps that mixes deterministic actions with AI checks, with versioning and reuse. Each step is a distinct action or QA decision point. Vision QA checks run alongside torque steps, OCR capture steps, video instruction actions, data capture actions and digital I/O outputs, and the API queues SuperPod motion sequences step by step. Steps branch on outcome.

A supervisor-authorized bypass records the step as Passing and a Next bypass as Failing, and incomplete records can be resumed or pulled for rework. Assemblies are edited with visual revision control, a revision lineage (inheritance tree) endpoint tracks versions, and approved releases can be rolled back or forward. Assemblies can be cloned for reuse. Rapta does not describe retry and fallback paths beyond rework.

Sourcerapta.ai/api-documentation, rapta.ai/knowledge-base/using-the-platform-training (20 May 2025), /knowledge-base/release-control; readread 2026-09-15

Every application runs on Sindri Forge, the company's execution platform. Forge runs a governed loop that ingests, validates, routes, executes and audits.

It assembles workflows from modular classification, extraction, reasoning, verification and orchestration components that are reused across applications, and it prioritizes deterministic execution paths and durable orchestration patterns. Sindri describes the approach as AI orchestrated and human governed.

It names no workflow designer, BPMN surface or customer facing builder, so the orchestration is Sindri's own and reaches customers as applications.

SourceSindri AI, sindri.ai/en/why-sindri/ai-technology and sindri.ai/enread 2026-09-08

Work reaches Rapta without a person starting it, through a machine signal. Input signals start assemblies when Supercoach sees a rising edge on a GPIO input from line automation, with set timing and Siemens IPC520a wiring, so a machine signal starts an assembly run with no person asking. The inline QA checks each step from the camera as work happens, not on request.

Reports can be scheduled through the API, with control over how often a report is emailed and who receives it, and supervisor lockout can email a supervisor when an operator is stuck. Digital I/O, barcode and RFID scanners and email are first-class channels. There are no webhooks, messaging tool or MES event subscriptions, and Rapta does not describe duplicate or race handling.

Sourcerapta.ai/knowledge-base/digital-i-o-signaling-and-connections (21 October 2023), /knowledge-base/supervisor-lockout-email-alerting, rapta.ai/api-documentation; readread 2026-09-15

Monitored inbound communications and detected exceptions start the work. The expediting assistant monitors supplier communications, predicts late risk and triggers escalation workflows. Procurement orchestration flags delays, and logistics visibility structures inbound unstructured PDFs and emails. Every run enters the loop through ingestion, before validation, routing, execution and audit. Work arrives on those events, not on request. Sindri names no schedule, webhook, inbound queue or channel as a separate way to start work.

SourceSindri AI, sindri.ai/en/solutions and /enread 2026-09-08

Checks are grounded in the customer's own parts, examples and assembly steps, but that knowledge enters by training a model, not through a retrieval layer. The platform learns proprietary assembly techniques from just a few correct and incorrect examples and generates the AI training set automatically, with a physics engine producing synthetic images of correct and defective parts. The API can create a new model for a specific assembly and start its training.

New customer knowledge cannot be added without retraining. A new assembly or revision enters by training a vision model, compiled into weights. There is no retrieval structure over the customer's own content that stays queryable between runs. Video work instructions are authored content shown to operators, not retrieval by the agent, and OCR capture of serial and batch numbers is data capture.

Sourcerapta.ai/ai-platform/how-it-works, rapta.ai/api-documentation; readread 2026-09-15

Agents ground their work in the customer's own data. A cross silo reasoning engine that connects siloed data is one of Forge's platform pillars. The workflows classify and extract from inbound PDFs and emails and reason over operational context that stays in the customer's tenant. Sindri names no index, graph or embeddings layer that persists, scales past the context window and stays queryable, so context appears to be assembled for each run.

SourceSindri AI, sindri.ai/en/why-sindri/ai-technology and /solutionsread 2026-09-08

What persists between runs is trained models and production records, not agent memory. Learned assembly models are trained weights, and work instructions, assembly records and operator metrics are the application's records. Learning across sessions is not a memory layer, and neither is the application's own database of business records.

The API can resume an incomplete assembly record or get assembly records for rework, the closest thing to run state, but these resume a production record, not a memory the agent keeps, and deleting that state deletes the traceability record. There is no session, conversation, workflow or long-term memory, so there is nothing to review, edit, delete or scope.

Sourcerapta.ai/api-documentation, rapta.ai/ai-platform/how-it-works; readread 2026-09-15

Forge lists explicit state management and durable workflow patterns under maintainability. They hold the state of each workflow run and the case being processed, as part of the application's data model and the durability of its orchestration. Sindri names no agent memory and gives no scope or lifetime for one.

SourceSindri AI, sindri.ai/en/why-sindri/ai-technologyread 2026-09-08

Checkpoints for each step, escalation and pause controls are built in and configured in Rapta's Setup tabs. With supervisor lockout (release 7.4.0), the Supervisor and Next buttons that bypass an AI-checked step can be password protected on any step type. A supervisor-authorized bypass records the step as Passing and a Next bypass as Failing, so an override is explicit.

Rapta can email a supervisor when an operator is stuck, and a Pause Live Operation control is listed in the knowledge base. Rapta does not route approvals into ticketing tools or explain why a checkpoint fired. Supervisor sign-off on new assembly revisions happens separately, under Release Control.

Sourcerapta.ai/knowledge-base/supervisor-lockout-email-alerting (5 November 2024), /knowledge-base/pause-live-operation (title in sitemap); readread 2026-09-15

Approval is built into the platform. A governance and approval hub is one of Forge's four platform pillars, and the platform routes approvals and executes approved actions. Policy controls in each workflow decide what can be automated, when people must approve, and how actions are executed and recorded.

Procurement orchestration enforces approval chains, and Sindri's security FAQ takes up whether approvals can be required for every action. The gate is set by policy and sits on the hub, between the route and execute steps of the loop. Sindri publishes no screenshot, documentation or escalation path for the hub.

SourceSindri AI, sindri.ai/en/why-sindri/ai-technology, /en/why-sindri/security-and-governance and /en/solutionsread 2026-09-08

Access controls are in place, but compliance rests on adherence to standards, not a held certification. Customer data is stored encrypted at rest with RBAC, the platform applies role-based access control, strong authentication mechanisms and regular access reviews, and a subset of telemetry is visible to team admins in the team log. Training requires supervisor-level authentication, and a supervisor sets a password in a Security tab. There is no SSO, SAML or SCIM.

Customer data stays on-premises and is never used to train other customers' models. On compliance, Rapta says it adheres to ISO 27001, 27002 and 27018 and follows ITAR-compliant practices, with products classified EAR99. No certificate, auditor, report or date is named, and no badge appears. Adherence to a standard is not an attestation, and no ISO 27001 certificate or SOC 2 report is published. EAR99 is an export classification, not a security attestation.

Sourcerapta.ai/security, rapta.ai/itar, rapta.ai/knowledge-base/supervisor-lockout-email-alerting; readread 2026-09-15

SSO is supported with optional two factor authentication and passkeys, depending on configuration. There is also a fully integrated identity management system, and policy controls set what can be automated and who must approve. Sindri names no SOC 2, ISO or other attestation, auditor or report. Its security brief is available on request, and automated InfoSec monitoring runs across a DevSecOps pipeline. No trust or security subdomain is published. Security material from the unrelated Sindri at sindri.app does not describe this vendor.

SourceSindri AI, sindri.ai/en/why-sindri/security-and-governance, probe recordedread 2026-09-08

Each Rapta assembly run produces a record of the system's QA verdicts for each step that a customer can inspect after the fact. The assembly report has photographic traceability and integrates serial, part, date-code and batch tracking, with each component's journey visually recorded and reports archived for long term storage to meet NTSB, FDA, DOE and NHTSA requirements.

The API exposes the record, with assembly records for each assembly, the recording for each record, endpoints to generate and retrieve a QA report for a specific record, and scheduled report delivery. Supervisor bypasses are recorded as Passing or Failing on the step, and a subset of telemetry and logs is visible to team admins in the team log. Operator metrics and Operator Analytics measure the human worker's performance, not the agent's. Rapta does not describe the model inputs and confidence behind each verdict, or SIEM export.

Sourcerapta.ai/ai-platform/ai-manufacturing-co-pilot, rapta.ai/api-documentation, rapta.ai/security; readread 2026-09-15

Every recommendation, approval and execution step is logged, reviewable and traceable for accountability, under what Sindri calls audit first execution. Forge treats auditability as traceable actions and approvals and audits every step, and Sindri promises a full audit trail where one is required. These records are Sindri's own, of what its agents recommended and did, and do not come from the customer's systems. Sindri names no log or trace view, retention period or export, and nothing that shows why an agent acted.

SourceSindri AI, sindri.ai/en/why-sindri/security-and-governance and /en/why-sindri/ai-technologyread 2026-09-08

Everything runs on an on-premises AI server, and customer data stays in the customer's facility. Customer data is stored and processed on that server and never goes to the public cloud. The station hardware is Siemens IPC520a industrial PCs with a Rapta Gateway at each workstation on a static IP, and the API exposes the gateway's network settings.

Data and processing are pinned to the customer's own site, so residency is answered by location, not by a region list. Software, models, source code and training data are developed and kept on U.S.-based systems by U.S. persons. There is one mode only, so there are no feature differences across modes. Rapta does not say who patches and upgrades on-site servers, and remote support access is optional.

Sourcerapta.ai/security, rapta.ai/itar, rapta.ai/api-documentation; readread 2026-09-15

Deployment happens inside the customer's cloud tenant and uses the customer's security controls, so operational context stays in the customer's environment to meet data residency and ownership requirements. An in tenant deployment model is one of Forge's platform pillars, and the platform supports private operation inside the tenant.

The privacy policy covers only the website and free trial services. Paid subscriptions come under a Master Service Agreement and Data Processing Agreement, and transfer to the US under SCCs applies to website data only. No hosting location for free trial environments is published.

SourceSindri AI, sindri.ai/en/why-sindri/security-and-governance, /why-sindri/ai-technology and /privacy-policyread 2026-09-08

Rapta names industry application solutions and worked examples, but every assembly is learned from the customer's own examples before it runs, so nothing is adopted ready-made. The named application solutions are Drones, Industrial Electronics, Laboratory Instruments, Precision Optics, Semiconductor, and Wiring & Panel Validation.

Application pages cover panel and wire inspection, kitting and part counting, multi-spectrum FOD inspection and TorqueGuardian, and worked examples cover realtime QA for electronics, bracket assembly and RMA, and panel wiring validation, with assemblies that can be cloned. These are industry starting points. Inspection routines that load for each assembly type and work instruction templates are the customer's own trained assemblies, not packaged assets. The application solutions are industry framings of one configurable platform, there is no asset a customer adopts and runs as is, and no prebuilt, ready to run inspection model is published.

Sourcerapta.ai/ai-platform/how-it-works, rapta.ai sitemap, rapta.ai/knowledge-base index; readread 2026-09-15

Four named products each have their own page. NC-Pro, also branded ANCR, handles non-conformance. BID-Pro handles bid compliance, CPMS Foundation supports capital project governance and DOC-Pro processes transmittals and documents. Sindri says it covers non-conformance resolution, bid compliance, transmittal processing and capital project governance today. Each is a whole application doing its own job on the shared Forge platform, and customers can adopt them one at a time. Workflows not yet built are delivered as custom work.

SourceSindri AI, sindri.ai/en and /en/solutionsread 2026-09-08

Model choice stays with Rapta, and no provider is named. Rapta says its AI models are trained exclusively on Rapta owned datasets, the API trains a model for each assembly and in a parts library, and a prompt based physics engine generates training images from text prompts.

No foundation model, provider or version is named for the vision models or the physics engine, no routing across models is disclosed, and there is no customer selection, bring-your-own-model or key path, including in the 224-path API, whose model endpoints create and train Rapta's models instead of choosing among them. Training a model on the customer's examples is customization, not model selection.

Sourcerapta.ai/security, rapta.ai/ai-platform/how-it-works, rapta.ai/api-documentation; readread 2026-09-15

The modular AI stack supports open weight models running privately inside the tenant, where applicable, to meet transparency and compliance needs. The platform can run that class of model, not only a single hosted provider. Sindri names no model selection for customers or admins, no providers or model list, and no routing control. AI training uses aggregated de-identified data and not content users submit.

SourceSindri AI, sindri.ai/en/why-sindri/ai-technologyread 2026-09-08

Rapta has an API for its own platform. The API documentation is a Swagger UI over an inline OpenAPI specification titled AI Supercoach, version 1.16.22, with 224 paths.

They include read-write endpoints for assemblies, steps, revisions and revision lineage, assembly records and QA reports, operators and operator metrics, model training (initiate, schedule, cancel), cameras, OCR capture variables, torque configuration, SuperPod motion control, and websocket channels for coaching updates, training status and device streams.

The platform is API-first, and there is an Example API Usage guide. There is no SDK, MCP server, or versioning and deprecation policy. Authentication is by operator credential login.

Sourcerapta.ai/api-documentation (Supercoach API 1.16.22), rapta.ai/ai-platform/enterprise-integrations; readread 2026-09-15

Outside callers have no published way in. Sindri names no API, SDK, developer portal, MCP server or agent card, its site navigation has no developer or docs entry, and there is no docs or dev subdomain.

Sindri Labs at sindri.app is a different company, a zero knowledge proving vendor, and its API, which it also calls Forge, comes with Python, JavaScript and Rust SDKs. The secure execution connectors reach out from the platform into the customer's legacy systems.

Sindri says its execution pattern works for workflows not yet built, and those are workflows Sindri builds, not ones a customer or outside caller drives through the platform.

Sourcesindri.ai/en/why-sindri/ai-technologyread 2026-09-08

Assembly changes wait for supervisor sign-off before they reach the floor. New revisions and changes are held for digital supervisor sign-off before updates are pushed to the factory floor, unapproved assemblies stay unpublished and invisible to operators, and revision control allows instant roll back or roll forward of an approved release. That is a quality gate in the release path of an assembly revision.

The gate is a person's approval, with no test run, score or pass rate to judge the revision against. Rapta does not describe testing the vision models with fixtures or datasets before production, or scoring output quality over time. Generating training sets automatically is how the model is built, not evaluation, and operator metrics measure the human operator, not the agent.

Sourcerapta.ai/knowledge-base/release-control (22 May 2024), rapta.ai/ai-platform/how-it-works, rapta.ai/api-documentation; readread 2026-09-15

No evaluation harness, scored test cases, release gate or controlled experiment after deployment is published, and Sindri names no way to measure the agents before or after a change. Sindri says deterministic execution paths make processes testable. The verification component in the Forge modular stack works with the validate step of the loop to check each output before it is routed. Sindri also says the product optimizes the customer's operations, and it offers a whitepaper.

SourceSindri AI, sindri.ai/en/why-sindri/security-and-governance and /en/why-sindri/ai-technologyread 2026-09-08

Cameras watch the physical work, and device interfaces control physical devices (SuperPod motion, torque wrenches, GPIO). None of that operates a graphical interface on the customer's computers, and camera based inspection has no screen to automate. Screen recording on the Rapta Station records the station's own display for support, and machine vision of parts is the product's quality function, not computer use.

Sourcerapta.ai/ai-platform/how-it-works, rapta.ai/api-documentation; readread 2026-09-15

Forge executes actions into legacy systems through secure execution connectors, not through an agent operating an interface. Sindri names no browser, desktop or remote computer control. Inbound PDFs and emails are structured by document extraction inside Sindri's own workflow.

SourceSindri AI, sindri.ai/en/why-sindri/ai-technology and /en/solutionsread 2026-09-08