agents.design.bio

A multi-agent assistant for biological design. Three specialized agents help you design, cultivate, and financially model bio-based materials — starting with bacterial cellulose. Each agent draws on its own knowledge base and tools; you can talk to them individually or let the system route your question automatically.

How to use

Talking to agents

Type your question in the message box. If your question has a clear topic (design, cultivation, finances), the system routes it to the right agent automatically.

To address a specific agent directly, type @ to open the mention menu and select an agent — or type @designer, @farmer, or @cfo followed by your question.

After each answer, two follow-up questions appear. Click any to continue the conversation in that direction.

To attach an image of a BC pellicle for analysis, click the 📎 button or drag and drop an image. Send it to @designer.

Use the ↺ button in the header to clear the conversation and start fresh.

The agents

Advises on bacterial cellulose material design — from cultivation parameters and post-processing to quality criteria and design applications. Can also analyze images of BC pellicles to estimate mechanical properties.

Answers questions about BC properties, drying methods, surface treatments, and plasticizers
Evaluates material against readiness levels (MR-1 through MR-3)
Recommends experiment designs (DoE) to improve target properties
Analyzes uploaded pellicle images: estimates tensile strength, elongation, stiffness, and uniformity — image analysis is powered by machine learning models hosted on Replicate
The app ships with a test model — download test_images.zip to try it out
Draws from an uploadable knowledge base of design criteria and research notes

Analyzes your production records — runs and treatment logs — to surface patterns, compare recipes, and identify what drives yield and quality. Connects directly to your Google Sheets data.

Queries and filters production data by date, recipe, or treatment
Ranks runs by yield, contamination, or any metric
Identifies which variables (temperature, media, inoculum) most influence yield
Detects anomalies and outliers in the dataset
Compares performance between time periods or recipe groups
Generates summary tables and trend analyses

Runs techno-economic scenarios for BC production. Given parameters like capacity, market mix, costs, and treatment methods, it computes revenue, EBITDA, net income, profit per kg, 5-year NPV, and payback period.

Models production at any scale with configurable capacity and utilization
Supports three market segments: fashion, automotive, and upholstery
Accounts for contamination and drying losses, quality grading, and treatment costs
Calculates NPV, ROI, and payback period
Runs sensitivity analyses ("what if I change X?")
Compares scenarios side by side (e.g. air dry vs. press dry)

Example @cfo reply with financial breakdown

Configuring settings

Accessing settings

Click the ⚙ icon in the top-right corner of the chat. The settings panel is password-protected — each private deployment can be configured with its own password and data sources.

For private deployments with custom data, knowledge base files, and branding, contact orkan@design.bio.

@designer

Upload .md knowledge base files containing design criteria, research notes, or material specs. Also configure the Replicate model version used for pellicle image analysis.

@farmer

Paste your Google Sheets URLs for the runs and treatments datasets. The app converts them to CSV automatically — the sheet must be publicly accessible or shared via link.

@cfo

Upload a .md file with YAML frontmatter to override the default techno-economic model parameters (capacity, prices, costs, etc.). If no file is uploaded, built-in defaults are used.

Settings page showing all three agent sections