> For the complete documentation index, see [llms.txt](https://alludium.gitbook.io/alludium-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://alludium.gitbook.io/alludium-docs/administration/3.-agent-builder/3.4-testing-and-deployment.md). # Testing and Deployment Treat agent testing like testing a production system — not an experiment. *** ### Testing Loop **1. Run dry runs on real past work**\ Feed the agent historical briefs, transcripts, or examples. Compare output to what your team actually produced. Would it have been useful? **2. Check for common failure modes**\ Missing context, overconfident conclusions, unsupported claims, tone mismatches. **3. Iterate in the Agent Configurator or Configuration sub-tabs**\ Adjust instructions based on what failed. You can make changes conversationally via the Agent Configurator chat or edit configuration directly in the relevant Configuration sub-tabs (Prompt, AI Model, Skills, etc.). Refine input specifications or output definitions. Update guardrails to prevent observed issues. **4. Test again**\ Run the same scenarios through the updated agent. Verify adjustments resolved the issues and test with new scenarios. **5. Repeat until outputs meet quality standards consistently**\ The goal is an agent that reliably supports your team's work rather than creating additional review burden. **Testing Mistakes to Avoid** ❌ Testing only happy path scenarios: Agents break on edge cases. ✅ Test failures: what happens with missing data, unclear requests, boundary conditions. ❌ Testing with toy examples: "Research Apple Inc" isn't realistic. ✅ Test with real work: use actual briefs, transcripts, and examples from your team. *** ### Pre-Deployment Checklist Before clicking Deploy, verify: * \[ ] Agent tested with diverse scenarios * \[ ] Outputs consistently meet quality standards (80%+ approval rate in testing) * \[ ] All required tools are connected and working * \[ ] Files are uploaded and linked correctly * \[ ] System instructions clearly define scope and guardrails * \[ ] Variables are properly configured (if used) * \[ ] Agent name and description are clear and accurate * \[ ] Category is appropriate for filtering * \[ ] You've tested failure scenarios (missing inputs, unclear requests) **If any item fails, continue testing and iteration before deploying.** *** ### **Deploying an Agent** Once you are satisfied with configuration: 1. Navigate to the **Deploy** tab in Agent Builder. 2. Click the **Deploy** button. 3. The system automatically provisions resources. 4. Wait for processing to complete. 5. The agent becomes available in My Agents with its updated configuration. Changes do not propagate to live use until deployment is complete. If behaviour appears unchanged after making edits, confirm you have navigated to the Deploy tab, clicked Deploy, and allowed processing time to finish. ### **Sharing After Deployment** Once an agent is deployed to Active state, two sharing options become available: sharing within your workspace and exporting to other workspaces. *Sharing with your workspace* When to share: the agent has had many successful invocations, quality is consistent and reliable, the workflow is valuable for multiple team members, and you're ready to maintain the agent for wider use. Sharing is managed from the **Share** tab in Agent Builder. The Share tab shows the current sharing status (Active or Inactive), connection scopes for each integration (e.g., "Alludium Platform -- User-configured"), and controls for sharing and unsharing. Sharing process: 1. Navigate to the **Share** tab in Agent Builder. 2. Review connection scopes — each integration is listed with its scope, showing which require individual user-configured connections (personal email, CRM) versus shared connections (API keys, workspace tools). 3. Click **Share with Workspace**. 4. The agent appears in My Agents for authorized users where that surface is available. A "Shared" badge confirms the agent is shared, along with an "Agent shared with workspace" confirmation. To stop sharing, click **Unshare from Workspace** on the Share tab. Your ongoing responsibilities as the owner: maintain agent quality, update configuration as needed, communicate changes to users, and monitor usage and feedback. See **Agent Sharing (in Workspace)** for detailed sharing procedures and permission models. *Exporting to other workspaces* Agents can also be exported as portable templates and imported into other workspaces, making it straightforward to distribute tested configurations to new teams or standardise setups across multiple workspaces. To export, click the **Export** button in the Agent Builder toolbar (next to the View button). This packages the agent's full configuration as a reusable YAML template file and downloads the file. To import into a different workspace, go to Agent Builder, create a new agent, select Import Agent YAML File, and import the downloaded file. Imported agents are automatically scoped to the target workspace. Existing connections in the agent will need configuration before using the agent. *** ### Next Steps **Templates** — Browse and deploy pre-built agent starting points for common workflows **Files** — Upload and manage reference documents that agents use for institutional knowledge **Integrations** — Browse external systems, available tools, and authenticated connections **Connections** — Set up authenticated access from an integration detail page **Tools** — Understand the actions agents can perform within connected integrations **Automations** — Schedule agents and workflows to run automatically --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://alludium.gitbook.io/alludium-docs/administration/3.-agent-builder/3.4-testing-and-deployment.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.