/

Agentforce Testing Copilot

1.0.0

Preview

Developer Tools

Agentforce Testing Copilot

Agentforce Testing Copilot

Agentforce Testing Copilot

A real QA team for your Agentforce agents

A real QA team for your Agentforce agents

Agentforce Testing Copilot

Agentforce Testing Copilot

Agentforce Testing Copilot

Give your Agentforce agents a real QA team, without hiring one. Agentforce Testing Copilot is an end-to-end testing platform inside Salesforce that discovers agent coverage, generates realistic AI test cases, executes them, judges results with an LLM, diagnoses failures, and helps fix well-understood issues automatically.

Key features

  • Agent discovery across org agents, topics, and actions

  • AI-generated test cases from actual agent configuration

  • Execution with LLM-based response judging

  • Root-cause analysis for failed tests

  • Autopilot fixing for well-understood failure patterns

  • Cross-channel visual testing for real browser-rendered conversations

Use cases

  • Regression testing for Agentforce prompt, topic, and action changes

  • QA coverage for production Agentforce agent portfolios

  • Cross-channel validation for Messaging for Web and Experience Cloud

  • Release confidence for teams scaling beyond manual chat testing

Industries

Cross-industry

Business need

Manual, unscalable conversation-by-conversation QA for AI agents. It automates functional, safety, jailbreak, and PII/compliance testing with LLM-judged evaluation

Requires

Agentforce-enabled org; AFTC_Admin/AFTC_Tester/AFTC_Viewer/AFTC_AutomatedProcess permission sets; Salesforce Connected App; at least one LLM provider API key; Node.js and Python runtimes

Compatible with

Agentforce Agent API; Messaging for Web, Experience Cloud, generic chat, Copilot sidebar; Anthropic/OpenAI/Google AI/Vertex/Azure/Ollama/Databricks via LiteLLM

Support level

Priority

Screenshots

See it in action

See it in action

See it in action

A closer look at the interface, dashboards, and day-to-day workflows.

A closer look at the interface, dashboards, and day-to-day workflows.

Product details

Everything you need to evaluate and deploy

Everything you need to evaluate and deploy

Everything you need to evaluate and deploy

From compatibility to security, here is what your team needs to greenlight this product and get it running.

From compatibility to security, here is what your team needs to greenlight this product and get it running.

Compatibility

API version

Salesforce API v65.0. Requires an Agentforce-enabled org; Experience Cloud and Lightning (LWC) channels supported; no Classic mentions.

Integrations required

Salesforce CLI; Salesforce org with Agentforce enabled; Node.js + Playwright (visual-test-service); Python 3.12 (testing-agent-team); Heroku/Postgres/Redis for auxiliary services; Salesforce Connected App (OAuth); at least one LLM provider API key (Anthropic/OpenAI/Google AI)

Setup time

Roughly half a day to 2 days for full setup, the core package deploys in minutes via sf CLI, but full functionality requires standing up two auxiliary services (Heroku/Postgres/Redis) plus OAuth Connected App configuration

App details

Languages

English

Package type

Metadata (Unmanaged)

Edition support

Enterprise, Unlimited, Developer (requires an Agentforce-enabled org)

Mobile compatible

Limited, 28-component Lightning dashboard suite designed for desktop administration, not mobile-optimized

Security

All Apex classes declared with sharing; no explicit CRUD/FLS enforcement calls found (relies on running-user permissions). Named Credentials send data to external LLM providers (Anthropic, OpenAI, Google AI) for judging. OAuth tokens encrypted at rest (Fernet) in the auxiliary service. Built-in PII detection/masking and compliance categorization (HIPAA/GDPR/PCI/COPPA/CCPA).