3rd Party LLM Connectors for TotalAgility
Download 3rd Party LLM Connectors for TotalAgility Overview
Features
Ready-to-run connector examples and templates for integrating third-party LLMs into TotalAgility as Custom Services:
-
AWS Bedrock — Amazon Nova models (AWS Lambda API): Route prompts and document data to Nova through your existing Bedrock deployment.
- Anthropic Claude models — Native API — connect directly to Anthropic's Claude API (Claude Opus, Sonnet, Haiku) without routing through a cloud provider
- Azure-hosted Claude models (Messages API, via Azure Foundry) — for orgs standardizing on Azure for model hosting and governance
-
Google Vertex/Studio — Gemini Pro models (Gemini API): Integrate Gemini Pro for multimodal document and image understanding.
-
Google Gemma models (via TotalAgility OpenAI-Compatible Connector): Run open-weight Gemma models through the same familiar connector pattern as OpenAI.
-
Qwen models (via TotalAgility OpenAI-Compatible Connector): Bring Alibaba's Qwen models into TotalAgility workflows without custom middleware.
-
xAI Grok models (xAI API): Connect directly to Grok for prompt-based tasks inside TotalAgility Cases and Agents.
Each example is a drop-in template — configure endpoint and credentials, then use the resulting Custom Service to power Custom LLM providers, Agents, or LLM steps in Workflows and Cases.
Benefits
Bring your own model, not just your own key. TotalAgility ships with native OpenAI integration, but many organizations standardize on Anthropic Claude, Google Gemini, Amazon Nova, or open-weight models like Qwen and Gemma — often for reasons like existing cloud commitments (AWS, Azure, GCP), data residency requirements, or model-specific capabilities. This connector suite lets you route TotalAgility's LLM calls through the provider your organization already trusts, without waiting on a native integration.
No added middleware. Connectors run as standard TotalAgility Custom Services — no separate integration layer to deploy, patch, or monitor. Credentials are stored securely in TotalAgility or a connected keystore, consistent with how you already manage other service credentials.
Same workflow surface, more model choice. Because these are Custom Services, they plug into the same Agents, Cases, and Workflow steps your team already builds with — swapping or adding a model provider doesn't mean rearchitecting the automation.
Technical Details
Inputs
Prompts, data, images and/or documents, including tiffs & PDF files.
Outputs
LLM responses (text).
Geographic Availability
Additional Information
FAQ
Does this require additional middleware or a separate integration layer?
No. The connectors run as standard TotalAgility Custom Services, so there's nothing extra to deploy or maintain outside of TotalAgility itself. Credentials are stored securely in TotalAgility or a connected keystore. (Exception - need Lambda function in between to be set in Bedrock for which script is provided in the repo)
Which TotalAgility version is required?
TotalAgility 2026.1 or later. The package is provided as a ready-to-import project for that release line.
Is Anthropic Claude supported, and how?
Yes — Claude models are supported via Azure Foundry using the Messages API. An example Custom Service connector is included in the package, along with connectors for AWS Bedrock (Nova), Google Studio (Gemini Pro), and OpenAI-compatible routing for Gemma and Qwen models.
Is there a cost to use these connectors?
The connector suite itself is free. You'll still pay your chosen provider (AWS, Azure, Google, xAI, etc.) for the underlying LLM usage under your own account with them.
Where do I get support if something isn't working?
Support is handled through the GitHub issues page for the project, or by emailing labs@tungstenautomation.com.“
For support, please use the issues feature on GitHub.
