Scale Donovan

What it does

Donovan takes in documents, knowledge bases, and external event data such as GDELT, and produces agent-driven analysis, reports, and answers with cited sources. Users configure agents with a system prompt, a knowledge base, and a chosen language model, then run them through a retrieval pipeline that includes a connector, retrieval, reranking, and the model.

People use Donovan to build agents without writing code, using a no-code Agent Factory, and to compare candidate models on metrics such as factual accuracy, hallucination rate, context recall, and latency. It is model agnostic and offers prebuilt agents such as a GDELT Agent, an Exhaustive Search agent, and a Mission Analyst agent. Traceability of reasoning steps is described as a feature.

Donovan suits public-sector and enterprise teams that need agent deployment in controlled environments. The page says it runs on classified and air-gapped networks, GovCloud at DISA IL4 and FedRAMP High, and Kubernetes-based containers. Access is by requesting a demo; pricing was not published in the supplied material, and model quality and reliability depend on the chosen model and data.

What it can do

• No-code Agent Factory for building custom agents from data sources, knowledge bases, models, and instructions
• Model comparison with metrics for factual accuracy, hallucination rate, context recall, and latency
• Agentic retrieval-augmented generation over uploaded documents, cloud sources, or APIs
• Prebuilt agents including GDELT Agent, Exhaustive Search, and Mission Analyst
• Traceable reasoning showing steps and resources an agent used
• Deployment on classified, air-gapped, GovCloud, and Kubernetes-containerized environments

Works with

• GDELT event data connector
• Scale GenAI Platform (SGP) integration for building agents

Example workflows

Build a document analysis agent

Inputs: uploaded documents and a system prompt. Steps: create a knowledge base, select a model, assign instructions in the Agent Factory, and run queries. Outputs: answers that cite the source documents.

Select a model for a workflow

Inputs: two candidate models and evaluation prompts. Steps: run the models in Compare Models and review factual accuracy, hallucination rate, context recall, and latency. Outputs: a comparison table to inform model choice.

Before you choose

• Deployment options depend on the chosen environment (classified, air-gapped, GovCloud, or Kubernetes); confirm availability with the vendor.
• Output quality depends on the selected model and data; the page reports metric examples but does not establish general performance.
• Pricing and procurement terms were not supplied and must be confirmed directly with the vendor.

Listed under

• Coding
• MCP and agent tools
• AI agents
• Defense
• Government
• Intelligence analysis
• Model evaluation
• No-code

Get started

The page directs visitors to request a demo.

1. Request a demo through the Get Started or Book a Demo link at scale.com/demo.
2. Define the agent's system prompt and connect a knowledge base or data source such as GDELT.
3. Select a model and run evaluations in Compare Models before deploying the agent.

Discuss the intended deployment, access requirements and commercial terms directly with Scale.

We couldn’t find published prices for Scale Donovan. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Donovan: Super Intelligence (SI) Agents for the Public Sector | Scale ↗scale.com
• Getting Started with Scale - Scale Documentation ↗scale.com
• Meet with our team | Scale AI ↗scale.com

Company website

scale.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.