Skip to content

Co-Engineer

← Home · Co-Engineer

The Protos Co-Engineer is a multi-agent AI system available across all features. It structures data, configures simulations, and surfaces connections — always with traceable sources. Depending on your task, the Co-Engineer hands off to specialised sub-agents best suited to help.

Important: Co-Engineer only surfaces information it can trace. It will not speculate or fill in gaps with assumptions. Every claim links back to a source you can inspect.

On This Page

What Co-Engineer Can Do

Capability Example
Schema authoring Create or update a schema based on your description — e.g. "Create a schema for tablet formulation experiments"
Data document creation Extract structured data from an uploaded file and create a data document following your schema
Canvas building Builds a simulation canvas — adds blocks, writes the code, and wires everything together
Knowledge surfacing Find relevant entries from the Knowledge Library as you work
Requirements parsing Parse an uploaded spec document into structured targets and constraints
Design comparison Compare design variants against requirements and surface gaps
Version history "What changed between v2 and v3 of this schema?" — browse and diff version history, check whether a canvas's dependencies are outdated, and restore a schema, canvas, or data document to a past version

How to Use It

The Co-Engineer panel is available from any screen via the chat icon in the top-right corner of the header bar.

Type in plain language:

  • "Create a schema for tablet formulation experiments with fields for particle size, binder type, and dissolution rate."
  • "Extract the key parameters from this test report and create a data document."
  • "Are there any prior experiments on this material in the Knowledge Library?"
  • "Build a canvas that takes my formulation data and calculates the adjusted capacity."

Co-Engineer responses always include source traces — click any claim to see where it came from.

You can attach files to any message — PDFs, documents, images, and more. The maximum file size per attachment is 100 MB.

Pro features: Some Co-Engineer capabilities require a Pro plan. If you're on a free plan you'll see a prompt to upgrade when you reach a Pro-only feature.

Multi-agent system

The Co-Engineer is built on a multi-agent architecture. Rather than a single assistant handling every task, it orchestrates specialised sub-agents and hands off to them mid-conversation depending on what you're doing.

Sub-agents

  • Co-Engineer — the generalist; handles open-ended questions and coordinates with the other agents
  • Knowledge — searches the Knowledge Library and the web to surface relevant information and prior work
  • Data — creates, edits, and validates data documents against your schemas
  • Simulation — builds and edits canvases in the Simulation Studio

Talking to indicator

The chat panel shows an agent badge with the current agent's identity icon and accent colour. Hover the badge to see a "Talking to" label. This updates automatically when the Co-Engineer hands off to a sub-agent.

Agent handoff

When the Co-Engineer delegates to a sub-agent, the chat shows a handoff message and the "Talking to" indicator updates. You don't need to do anything — the handoff happens automatically based on what you asked.

Slash commands

Typing / in the chat composer opens a command picker. Available commands:

  • /agent — switch to a specific sub-agent by name
  • /help — get context-aware help from the docs
  • /feedback — submit feedback about Co-Engineer

Versioning skill

Ask about version history, whether a canvas's dependencies are current, or to restore something to an earlier state, and the Co-Engineer loads its versioning tools automatically — there's nothing to turn on. See Versioning for what it can and can't do (model and knowledge-document versions can only be browsed, not restored, this way).

MCP servers

An MCP button in the lower-left of the chat composer lets you toggle individual MCP servers on or off for the current conversation. You can also check Default in new conversations per server so your preferred setup carries over automatically. If no servers are configured yet, the popover links you to MCP Connections to add one.

Sharing a chat session

Co-Engineer sessions can be shared with community members. Open a chat session, click Share, and assign Editor or Viewer access to individuals, teams, or the whole community. Shared sessions appear under Shared with me in the chat session list. See Collaboration & Sharing for details.

Co-Engineer and Your Data

Co-Engineer can take direct actions in your project — creating schemas, data documents, canvases, and requirements — based on what you ask it. These actions happen through the chat; the Co-Engineer will tell you what it created or changed.

You can review all Co-Engineer sessions for a project by opening the Co-Engineer panel and browsing past sessions.

How Co-Engineer Gets Smarter

Co-Engineer's quality depends on the richness of your workspace:

  • Models: The more models registered and documented, the better Co-Engineer's simulation configuration suggestions.
  • Knowledge Library: The richer the library, the more relevant connections Co-Engineer can surface as you design.
  • Schema quality: Well-defined schemas with units and descriptions give Co-Engineer more signal when extracting data from files and creating documents.

Investing in these foundations makes Co-Engineer progressively more useful over time.

See Also

  • Knowledge Library — Co-Engineer draws on this to surface relevant prior work
  • Models — Co-Engineer uses this for simulation configuration recommendations
  • Schemas — Co-Engineer can create and update schemas based on your description
  • Simulation Studio — Co-Engineer can recommend model and input configurations
  • MCP Connections — connect external tools the Co-Engineer can use in chat
  • Versioning — what the Co-Engineer's version tools can browse, diff, and restore

← Back to Home