Cosmon vs Synera (2026)

Nexus writes into named engineering applications, connecting to SolidWorks, Ansys, COMSOL, Abaqus and Star-CCM+. The SDK adds Fluent and Ansys Mechanical, the change log adds Fusion 360 and Catia V5, NX is also named, and Jama and Cameo are named on the requirements side, eleven commercial applications across CAD, CAE solvers and requirements or MBSE tooling. The writes carry weight, with automated geometry edits, parametric variations, dimensioned drawings with GD&T and BOMs authored natively inside SolidWorks, mesh generation, material and boundary condition assignment, simulation runs and PDF export.

Changing geometry in a customer's CAD model is a write into their system of record, and a riskier one than a CRM update. The integrations are versioned and maintained. Release 0.1.22 moved the SolidWorks FEA agent into production, let Fusion 360 author drawings and export PDFs and expanded validated workflows for CATIA V5, alongside corporate proxy support, and a COMSOL AI agent shipped separately. There is no integrations page, supported version matrix or connector catalog.

Sourcedocs.cosmon.com front page, batch22a and batch19a change entries, pypi.org cosmon-agent-sdkread 2026-09-13

Agents can read and write in 80+ named engineering tools, call web APIs with OAuth and use remote MCP servers as tools.

Synera connects 80+ CAD, meshing, analysis and productivity tools with 1000+ nodes, including read-write control of CAD applications such as CATIA, NX and SolidWorks to generate, modify and optimize geometries, FEA meshers and solvers such as HyperWorks, ANSYS and Abaqus, a Teamcenter connector for working with a Siemens Teamcenter environment, ODBC database queries, and web API nodes that call any REST API and authenticate with OAuth.

Release 26.07 lets customers plug any existing remote MCP server straight into an agent or Supervisor node as a tool. Connectors are native plus partner-built, with 20+ software partners in the marketplace, and custom nodes and Python, C#, MATLAB and TCL scripts add custom tools. Credential scoping, rotation and revocation are not publicly described.

Sourcesynera.ai/platform, /webinar/product-release-26-07-quick-quilla; portal.synera.io FAQ Integrating Synera with External Software; readread 2026-09-16

Nexus orchestrates across tool boundaries, which is the product's whole thesis. Cosmon places engineering AI above the individual platforms, automating not just tasks inside each silo but the transitions between them, from requirements tool to CAD to solver to PLM. A closed loop runs the full chain, from a requirement change in SysML through CAD updates and simulation reruns under governance to a traceable review package.

The steps are set out one by one, from turning natural language requirements into exact boundary conditions and pass or fail criteria, through geometry cleanup for meshing that understands the physics and keeps critical structural features, material and boundary condition assignment and solver runs, to results interpretation and report drafting.

In the SDK, a single client.run() call such as "Run a static study on bracket.sldprt" emits a sequence of typed ToolCall events, and runs can work for many minutes. Agents are addressed by name (agent="solidworks"), so work is routed to specialized agents, and a SolidWorks FEA agent and a COMSOL agent moved into production separately.

Because the SDK lets a customer batch, script and embed Nexus in their own pipelines, the customer composes the sequence and its surrounding logic in their own Python instead of accepting a fixed pipeline.

Sourcedocs.cosmon.com CAD and CAE sections, pypi.org cosmon-agent-sdk run model, cosmon.com Engineering OS positioningread 2026-09-13

Deterministic workflow nodes and autonomous agent steps mix in both directions under coordinating agents. Engineers build low-code workflows that are then made available as tools for agents, with management agents coordinating tasks while engineering agents focus on specialized processes. A coordinator agent manages the full workflow, passing context between specialized agents, maintaining state across the process and handling exceptions when input data is incomplete, and a Supervisor validates output before each handoff.

Release 26.07 adds one entry point that brings agents into both multi-agent conversations and rule-based workflows. Workflows are saved as templates or new nodes in individual or shared team libraries, publishing errors are shown as the workflow is built, and workflow errors can be inspected on the Run server since release 25.11. Explicit branching and fallback-path configuration for agent steps is not described publicly.

Sourcesynera.ai/ai-agents, /news/ai-agent-orchestration-engineering-use-cases, /webinar/product-release-26-07-quick-quilla, /platform; read; synera.ai/news/ai-agent-orchestration-engineering-use-casesread 2026-09-16

Nexus can run unattended from a customer's pipeline, but it is not confirmed that it starts on its own from an engineering event. Cosmon's insights pages describe change carried through the toolchain by events, where a requirement change in SysML can autonomously trigger CAD updates, rerun simulations under strict governance and generate a traceable review package.

There are two ways in, both started by the caller, working with the agent in the app or driving the same agent from the customer's own Python with the Agent SDK. Unattended runs are supported, and the SDK exists to batch, script and embed Nexus in pipelines, so a pipeline can call Nexus on a commit or a nightly build with no person involved, though the scheduling belongs to the customer's CI, not to Cosmon. Nothing shows Cosmon itself watching a requirements tool or PLM revision and firing.

Sourcedocs.cosmon.com invocation paths, pypi.org cosmon-agent-sdk unattended moderead 2026-09-13

Every entry point to Synera's agents is invoked by a person or by another system calling in.

Agents are reached through multi-agent conversations with a Supervisor, through Synera Run workflows an engineer opens by link and feeds input files, through the built-in Synera Run MCP server that external agents such as Copilot Studio, Claude and ChatGPT call, and through Synera Headless, which external software can start from a command line (syneraheadless.exe execute).

No schedule, webhook, or PLM or file event starts an agent without someone initiating it, and no requirements or design change trigger is described. Headless can be woken by the customer's own scheduler, but that is the customer's mechanism, not Synera's. Synera says its agents work continuously around the clock.

Sourcesynera.ai/webinar/product-release-26-07-quick-quilla, /ai-agents; portal.synera.io FAQ (Synera Headless, Integrating Synera with External Software); readread 2026-09-16

Nexus indexes engineering artifacts, not just documents. Cosmon describes indexing the actual physics and geometry of files so engineers can query prior work by concept, for example past thermal analyses of battery packs with a particular cooling geometry, which is retrieval keyed on geometry and physics against a stored corpus.

Nexus works directly inside the customer's engineering software, understanding designs and running simulations, and Cosmon names multimodal geometry technology that understands CAD structure, physics and design intent. In the SDK, "What's the max von Mises in bracket.sldprt?" is answered from a model file.

Grounding comes from the customer's systems of record, geometry in CAD files, results in solver output and requirements in Jama or Cameo, with PLM metadata interpreted so certification simulations run only on the correct released revision. Natural language requirements become exact boundary conditions and pass or fail criteria, so the numbers come from the requirement document, not the model.

Outputs are grounded in the customer's engineering sources of truth and produce a traceable review package, so a result can be traced back to the requirement and revision it rests on. Cosmon does not disclose its index technology, embedding approach or retrieval mechanism.

Sourcedocs.cosmon.com front page, cosmon.com multimodal geometry section, pypi.org cosmon-agent-sdk worked exampleread 2026-09-13

Two persistent retrieval structures sit over the customer's own content, Knowledge Spaces and geometry similarity search. Release 26.03 Phenomenal Phoenix introduced Knowledge Spaces, which let agents accumulate and apply expert decisions over time, taught through a Remember this interaction that stores, manages and reuses that knowledge. The same release added Part Comparison nodes that turn CAD geometry into a 100-dimensional fingerprint and search an entire part database for similar designs.

Release 26.07 adds an Advanced AI add-in that finds similar parts across a product lifecycle management (PLM) system or local files, and in the RFQ case a geometry agent searches the CAD library for comparable parts. Both persist, stay queryable between runs and take new knowledge without retraining. Workflows are tools the agent runs, not knowledge it retrieves. Document ingestion and refresh, retrieval visibility, citation of sources and stale entry handling are not publicly described.

Sourcesynera.ai/webinar/product-release-26-03-phenomenal-phoenix, /webinar/product-release-26-07-quick-quilla, /news/ai-agent-orchestration-engineering-use-cases; readread 2026-09-16

Indexing and recalling past analyses as institutional knowledge is a retrieval corpus about the customer's designs, which is knowledge, not state the agent keeps about its own work. No agent carries state across runs.

There is no memory scope by user, project or agent, no published lifetime or retention, no expiry or purge path, no way to read or write accumulated state, and no recall of a previous run as distinct from a previous file. The SDK points the same way.

A run is created by Client.run(), yields typed events and returns an answer, nothing in the client saves, names, resumes or refers to an earlier run, and attended or unattended mode is set for each client, not carried forward.

Sourcedocs.cosmon.com, pypi.org cosmon-agent-sdk run modelread 2026-09-13

Conversation state carries through a session. Each message sends the system prompt, tool descriptions and the conversation history to the model, history grows through a session, and starting a new session creates a fresh conversation, while the coordinator agent maintains state across a multi-agent run. That state disappears when the session ends.

No per-user or per-agent memory with a stated scope and lifetime is shown, so there is no memory store to review, edit, delete or scope. Knowledge Spaces is a store of expert decisions that agents query, and a workflow library is the application's own data.

Sourceportal.synera.io FAQ How does token usage work in MAS; synera.ai/news/ai-agent-orchestration-engineering-use-cases, /webinar/product-release-26-03-phenomenal-phoenix; readread 2026-09-16

Oversight is a setting, and the interactive path is review and approve. In the SDK, whether a run is attended is set when the client is built, by whether handlers are supplied, so a caller decides in code whether the agent may proceed alone or must call back to a person. Cosmon sets the gate for each workflow.

Nexus diagnoses solver issues and proposes fixes that a person reviews and approves before any change is made, flags manufacturability, assembly risks and cost issues for the engineer to decide on, and interprets simulation results and drafts reports for review and sign off.

Engineers grant permission for actions, and even the closed loop case, where a requirement change carries through CAD updates and reruns simulations under strict governance, runs without human intervention only until final sign off. Every AI decision can also be reversed. Cosmon frames it as engineers deciding the direction while Nexus executes.

There is no approval queue, no risk tiering by action, no record of who approved what, and no policy at the organization level that forces attended mode, so whoever writes the client chooses, and an engineer can opt out of their own oversight.

Sourcepypi.org cosmon-agent-sdk attended versus unattended, cosmon webinar workflow descriptions, docs.cosmon.comread 2026-09-13

The engineer's role shifts to reviewer while the agents handle cross-tool execution, and the engineer retains judgment over the results. In the RFQ build a requirements agent flags ambiguities for engineer review, and in aerospace work the engineer directs the work. No configurable sign off is shown, and where approvals can be inserted, whether at node, workflow or policy layer, who signs off and what happens when nobody does are not described. Product documentation at portal.synera.io/docs and /dev-docs sits behind a login.

Sourcesynera.ai/news/ai-agent-orchestration-engineering-use-cases, /news/agentic-ai-for-engineering-in-aerospace-and-defense; portal.synera.io API probes; readread 2026-09-16

There is no attestation, but a thin access model for each user exists. There is no SOC 2, ISO 27001 or GDPR statement, no named auditor or report and no trust center. Local deployment that keeps IP inside the network is a deployment property, interpreting PLM revision rules is integration with the customer's PLM, and keeping designs confidential is a promise, not a mechanism.

The SDK's identity model is a real control customers use. Each user works with their own login and credits, code never holds a key, capability is bound to the signed in installed app instead of a shared secret, and Developer access is a setting a user must turn on before the local MCP server can be driven. There is no SSO, SAML, RBAC, SCIM or admin console.

Sourcedocs.cosmon.com, pypi.org cosmon-agent-sdk authentication modelread 2026-09-13

Synera holds TISAX Level 2 and brokers single sign-on to the customer's identity provider. TISAX is the automotive information-security assessment derived from ISO/IEC 27001, independently audited and verifiable on the ENX portal under Scope ID S007M3 and Assessment ID AX6WMT.

The Synera Identity Manager (IDM, Keycloak) is an SSO broker between the customer's identity provider, such as Entra ID, Active Directory, SAML or OIDC, and Synera components, and Run server publishing is permission-controlled. No canvas data is sent with analytics, only usage statistics and interface interactions. TISAX results are not public by design, and role based access across tools, knowledge and model usage is not publicly described.

Sourcesynera.ai/news/synera-achieved-tisax-level-2-certification; portal.synera.io FAQ What are the 5 components of the Synera server stack, Since Synera is an open platform what features are there for security; readread 2026-09-16

Each run exposes a step by step trace, and agent actions can be undone. Cosmon says every AI decision stays visible, traceable and fully reversible, and the SDK supplies the mechanism, since iterating a Run yields typed ToolCall and TextChunk events, with the example printing each tool step as it happens (print(f"[{event.summary}]")) while the answer streams token by token. A caller can watch which tools the agent used, in order, with a summary of each, while the run is in progress.

Reversibility means an agent action on a CAD model can be undone, which requires the platform to know what it changed, and on a product that edits geometry in a customer's system of record, that is the auditability that matters. A closed loop change produces a traceable review package that keeps the record of what was done for sign off. Coverage reports, solver diagnostics and manufacturability findings describe the customer's part, not the agent. There is no retention period, saved run history a customer can browse later, or admin audit log across users, and the trace is available to whoever calls a run.

Sourcepypi.org cosmon-agent-sdk event streaming, cosmon.com traceability and reversibility claimsread 2026-09-13

The Supervisor agent narrates every step of a run across several agents, from the inputs taken from the engineer to the outputs delivered to each specialized agent, and every geometry change, simulation result and calculation remains visible and accessible in the Synera UI.

Release 26.07 lets users inspect 3D outputs an agent just produced directly in the chat, release 25.11 adds Workflow Diagnostics to inspect workflow errors directly on the server, and customers can collect MAS server log files for support.

Audit logs separate from runtime traces, export to monitoring or SIEM, and retention are not described, and there is no built-in token tracking.

Sourcesynera.ai/news/agentic-engineering-blueprint, /webinar/product-release-26-07-quick-quilla, /news/release-webinar-25-11-outstanding-odysseus; portal.synera.io MAS FAQ; readread 2026-09-16

The customer chooses where Nexus runs, in the cloud or locally, so proprietary engineering IP never leaves the customer's network. Nexus ships as an installed desktop application, with release notes on a desktop app subdomain, an SDK that needs an installed, signed in Cosmon app, and capability that comes from the CAD software on the engineer's machine. Software running on the engineer's own workstation against locally installed CAD puts the software where the customer is, which is different from a residency promise about a vendor cloud.

Release 0.1.22 added corporate proxy support for managed networks, which a vendor builds only because customers run the product inside controlled corporate networks. Geometry and simulation results from before a product's release are among a hardware company's most closely held assets. No region, region list, country or residency commitment is named for the cloud option, and nothing describes what local deployment involves.

Sourcedocs.cosmon.com, pypi.org cosmon-agent-sdk installed-app model, batch22a change entry on corporate proxy supportread 2026-09-13

Synera installs on the customer's own infrastructure, with an AWS option for the agent solution. The server stack has a Desktop Client installed on each user's machine, a Token License Server (RLM) and an optional Synera Server that exposes the Run Daemon and MAS. It also has an Identity Manager and PostgreSQL databases, which customers who already run their own PostgreSQL server can simply create there, and consumer users reach the customer's Synera Server URL in a browser.

Automation experts host their engineering workflows on an on-premise Synera Run server, installers and rollback are available for Windows 10/11, and the agent solution is deployable on AWS. On-prem, hybrid, local runtime and self-hosting are available now, and the customer installs upgrades and rollbacks from the portal. Feature differences between self-hosted and AWS deployment are described only for Run versus local machine, and no Synera-hosted region list is published.

Sourceportal.synera.io FAQ server stack; synera.ai/platform, /releases, /ai-agents; readread 2026-09-16

Named agents for each tool can be selected, but no catalog exists. Agents are addressed by name in the SDK (client.run("...", agent="solidworks")), so a customer chooses which agent does the work, and the change log names them as separate shipping units, with the SolidWorks FEA agent moved into production and a COMSOL AI agent released separately.

The capabilities fall into families, CAD (modeling, drawings, DFMA, tolerance stack up and reverse engineering) and CAE (simulation preprocessing, troubleshooting and postprocessing). There is no agent gallery, lasting page for each agent, template library, enable or install action, or entitlement language, and documentation of what the product can do is not a set a customer picks from. Custom build outs are professional services.

Sourcedocs.cosmon.com CAD and CAE sections, pypi.org cosmon-agent-sdk agent parameter, batch22a and batch19a change entriesread 2026-09-13

Workflow templates come as a browsable catalog of downloads. The Synera Marketplace lists 189 templates, each a downloadable .syn workflow tagged by application and industry, such as Advanced FEA, Advanced FEA with Altair HyperMesh and OptiStruct, Advanced LS-DYNA Simulation Modification and AM Line Supports, and Synera offers pre-built templates for different industries and applications.

Each template runs as a whole workflow when opened, and templates are editable in the visual editor and can be saved as custom nodes. The catalog mixes production-grade simulation workflows with small utilities such as Alphabetical Sorting, no prebuilt agent teams are cataloged, and customer references are not tied to templates.

Sourceportal.synera.io/marketplace and /api/marketplace/v2/templates; synera.ai/platform; readread 2026-09-16

The customer does not choose the model, and none is named. Cosmon describes multimodal geometry technology that understands CAD structure, physics and design intent, trained on thousands of synthetic engineering problems across domains, which points to proprietary geometry models, and a vendor that trains its own models has a coherent reason to fix the stack.

There is no model selector, admin control, model setting per agent or run, disclosed routing, or bring your own model or key path, and the SDK's run signature takes a prompt and an agent name with no model argument. No provider, model family or version is named, and Cosmon contrasts Nexus with generic agents built on LLMs without saying what sits underneath. A hardware company sending geometry before release cannot learn which model reads it.

Sourcecosmon.com multimodal geometry section, docs.cosmon.com, pypi.org cosmon-agent-sdk run signatureread 2026-09-13

Customers assign models per agent on their own model deployments. In Managed AI Services (MAS), one agent may be assigned a model with a lower token limit while other agents use models with higher limits, and customers can deploy a model with a higher token limit for the Supervisor in their model deployment settings, for example Azure OpenAI, with rate limits set on their own LLM deployments, such as Azure AI Foundry, by their IT administrator.

Synera advises choosing models with larger context windows, and release 26.07 adds GPT-5-class model support. Model choice is customer and admin controlled, per agent, on the customer's own providers and keys. No provider list beyond the Azure examples is given, and model policy by workspace or action, benchmarking and fallback policy are not described beyond automatic retry with back-off on rate limits.

Sourceportal.synera.io FAQ entries: Why does the Supervisor in MAS sometimes respond slowly or get stuck; How does token usage work in MAS; synera.ai/webinar/product-release-26-07-quick-quilla; readread 2026-09-16

Cosmon publishes a Python SDK for driving its agent. Agent SDK is one of five top level cards in the documentation, offering to drive Nexus from Python to batch, script and embed it in pipelines, with its own section at /sdk/overview, and the product can be used in two ways, interactively in the app or driven from the customer's own Python.

The package is cosmon-agent-sdk on PyPI, published from support@cosmon.com, with four releases through 0.1.3, installed with pip install cosmon-agent-sdk. Its client surface includes Client(), client.run(prompt, agent=...), a Run object with .text(), typed ToolCall and TextChunk events for streaming, attended or unattended mode set by whether handlers are supplied, and a cosmon doctor diagnostic command.

Under the hood the SDK is a small MCP client that finds the running app, connects to the local MCP server it exposes and drives the agent through one run tool, once the user turns on Settings > Developer access. The caller's code brings the request and the installed app brings the capability (the user's login, credits and local CAD software), so the code never holds a key. The SDK is marked alpha at 0.x, with breaking changes expected.

Sourcedocs.cosmon.com Agent SDK section, pypi.org cosmon-agent-sdk published by support@cosmon.comread 2026-09-13

Synera serves its workflows to outside agents through a built-in MCP server and to outside software through a command-line runner. Release 26.07 Quick Quilla added the built-in Synera Run MCP server, which lets external agents such as Copilot Studio, Claude and ChatGPT execute the engineering workflows a company's experts already built, turning a Synera workflow into a tool an external agent can execute from Codex, Cursor and VS Code as well.

Synera Headless runs any workflow from a command line (syneraheadless.exe execute --file --jsoninput --output), so external software can start Synera and read its results, and developer documentation sits at portal.synera.io/dev-docs. Synera both consumes and serves MCP, and custom nodes and Python, C#, MATLAB and TCL scripts extend the platform. No REST API or SDK reference is publicly readable, since /dev-docs requires a login, and authentication and versioning of the MCP server are not publicly described.

Sourcesynera.ai/webinar/product-release-26-07-quick-quilla, /releases, /platform; portal.synera.io FAQ Synera Headless; readread 2026-09-16

Nothing tests the agent itself. Cosmon's testing is about the customer's design, since running a simulation to prove a bracket survives a load, flagging manufacturability and draft issues, checking tolerance stack up and diagnosing a solver that failed to converge all evaluate the customer's part, not Nexus. Design validation, issue detection and holistic optimization are not agent evaluation.

There is no harness for agent output, scored cases, regression suite over generated geometry, golden set, judge, comparison between agent configurations or accuracy figure for the agent's own work. Cosmon says Nexus is trained on thousands of synthetic engineering problems across domains and delivers best in class outcomes where generic agents built on LLMs fall short, with no benchmark, comparator or number. Being able to undo a decision is control after the fact, not a test before it.

Sourcecosmon.com product claims, docs.cosmon.comread 2026-09-13

A workflow is checked before it is released to users. Release 25.11 adds a new publishing checklist for Synera Run, and possible publishing errors are shown dynamically as the workflow is built. The MAS Supervisor retries automatically with exponential back-off when an LLM rate limit is hit.

The publishing checklist checks whether a workflow will publish and run, not what the agent produced, and testing with fixtures or datasets before production and scoring output quality over time are not described. Simulation and validation loops in Synera test the customer's part design, not the agent workflow.

Sourcesynera.ai/news/release-webinar-25-11-outstanding-odysseus; portal.synera.io FAQ Why does the Supervisor in MAS sometimes respond slowly or get stuck; readread 2026-09-16

Nexus drives desktop CAD on the engineer's machine, but nothing shows it doing so through the screen. It ships as a desktop application that works directly inside the customer's engineering software, connecting to SolidWorks, Ansys, COMSOL, Abaqus and Star-CCM+, the capability is local (the installed app brings the user's login, credits and local CAD software), and the change log adds a Nexus Connector and corporate proxy support for managed networks. An agent driving locally installed desktop CAD comes close to local computer control, but the interface decides it.

Every CAD and CAE application named, including SolidWorks, Ansys, Abaqus, COMSOL, Fluent, Star-CCM+, Fusion 360 and Catia, exposes a scripting or automation API, the ordinary way software drives them, and the published outputs are what those APIs produce, fully dimensioned engineering drawings, multiple views, sections, GD&T and BOMs inside SolidWorks, mesh and boundary condition setup, and PDF export. Nothing from Cosmon describes screen capture, cursor control, pixel reading or targeting interface elements.

Sourcedocs.cosmon.com front page, pypi.org cosmon-agent-sdk, cosmon.com SolidWorks drawing outputsread 2026-09-13

Engineering applications are driven through their APIs, though one marketplace add-in may automate graphical interfaces. Synera's CAD control of CATIA, NX and SolidWorks and the Run any CAD add-in, which automates any CAD application offering an API, both drive applications through their APIs, not through the screen.

The RPA by stone-dev add-in in Synera's own marketplace covers applications that cannot be connected through standard add-ins or Python, Run External Program or Web Request nodes. It may be GUI automation shipped in Synera's catalog, but the marketplace search and the add-in page sit behind a login, so whether it drives a visual interface is not published.

Sourceportal.synera.io FAQ Integrating Synera with External Software; synera.ai/platform; readread 2026-09-16