Workspace Integrations

Configure AI providers, Git connections, and notification channels per workspace with encrypted secret storage and workspace-level overrides.

TestMesh supports workspace-scoped integrations, letting each workspace use its own AI provider, Git connection, or notification channel. Workspace integrations override global defaults, and secrets are encrypted at rest using AES-256-GCM.

Integration Types

AI, Git, and Notification providers

Provider Resolution

How TestMesh picks which provider to use

AI Provider Routing

Per-agent LLM provider overrides

Dashboard Setup

Configure integrations in the UI

Integration Types

Type	Providers	Purpose
AI Provider	OpenAI, Anthropic, Local (Ollama)	LLM-powered test generation, analysis, self-healing
Git	GitHub, GitLab, Gitea	Webhooks, diff analysis, PR write-back
Notification	Slack	Alerts on execution failures, schedule triggers

Provider Resolution

When TestMesh needs a provider (e.g., an AI model to analyze a failure), it resolves in this order:

1. Workspace-specific integration  →  found? use it
2. Global integration              →  found? use it
3. Environment variable fallback   →  ANTHROPIC_API_KEY / OPENAI_API_KEY

This means you can set a global Anthropic integration for all workspaces, then override with OpenAI for a specific workspace that needs GPT-4 — without changing environment variables or restarting the server.

Managing Integrations

Create a Workspace Integration

curl -X POST http://localhost:5016/api/v1/workspaces/$WORKSPACE_ID/integrations \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Team OpenAI",
    "type": "ai_provider",
    "provider": "openai",
    "config": {
      "model": "gpt-4o",
      "temperature": 0.7,
      "max_tokens": 8000
    },
    "secrets": {
      "api_key": "sk-xxxxx"
    }
  }'

Create a Global Integration

Omit workspace_id or use the admin endpoint:

curl -X POST http://localhost:5016/api/v1/admin/integrations \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Default Anthropic",
    "type": "ai_provider",
    "provider": "anthropic",
    "config": {
      "model": "claude-sonnet-4-5-20250514"
    },
    "secrets": {
      "api_key": "sk-ant-xxxxx"
    }
  }'

Test a Connection

curl -X POST http://localhost:5016/api/v1/admin/integrations/$INTEGRATION_ID/test

Returns the test result and updates last_test_at / last_test_status on the integration.

List Integrations

# Workspace integrations (includes global fallbacks)
curl http://localhost:5016/api/v1/workspaces/$WORKSPACE_ID/integrations

# Global integrations only
curl http://localhost:5016/api/v1/admin/integrations

Secret Storage

Integration secrets (API keys, tokens, webhook secrets) are encrypted using AES-256-GCM with per-secret nonces before storage. They are never returned in API responses — only a "configured": true flag is shown.

{
  "id": "integration-uuid",
  "name": "Team OpenAI",
  "type": "ai_provider",
  "provider": "openai",
  "config": { "model": "gpt-4o" },
  "secrets": { "api_key": "configured" }
}

To rotate a secret, update the integration with the new value:

curl -X PATCH http://localhost:5016/api/v1/admin/integrations/$INTEGRATION_ID \
  -H "Content-Type: application/json" \
  -d '{
    "secrets": {
      "api_key": "sk-new-key-xxxxx"
    }
  }'

AI Provider Routing

Beyond workspace-level defaults, you can route specific AI agents to specific providers. For example, use Anthropic for test generation (better at structured output) and OpenAI for embeddings (lower cost).

Configure Per-Agent Overrides

curl -X PUT http://localhost:5016/api/v1/workspaces/$WORKSPACE_ID/ai-config \
  -H "Content-Type: application/json" \
  -d '{
    "default_provider": "anthropic-integration-uuid",
    "agent_overrides": [
      {
        "agent_name": "coverage",
        "integration_id": "openai-integration-uuid"
      },
      {
        "agent_name": "generation",
        "integration_id": "anthropic-integration-uuid"
      }
    ]
  }'

Resolution Chain

For each agent call, TestMesh resolves the provider:

1. Agent-specific override     →  coverage → OpenAI
2. Workspace default provider  →  Anthropic
3. Global default              →  whatever is configured globally
4. Environment variable        →  ANTHROPIC_API_KEY / OPENAI_API_KEY

Available Agents

Agent	Purpose
`coverage`	Analyze test coverage gaps
`impact`	Determine blast radius of changes
`diagnosis`	Root-cause analysis for failures
`repair`	Generate fix suggestions
`flakiness`	Detect flaky test patterns
`generation`	Generate new test flows
`watch`	Monitor and adapt tests
`scheduler_optimizer`	Optimize test scheduling
`orchestrator`	Coordinate multi-agent workflows

Dashboard

Integrations Settings

Navigate to Settings → Integrations in the workspace dashboard. The page shows three tabs:

AI Providers — Add OpenAI, Anthropic, or Local providers
Git — Connect GitHub, GitLab, or Gitea
Notifications — Configure Slack channels

Each integration shows its status (active/disabled/error), last test result, and a Test Connection button.

AI Providers Settings

Navigate to Settings → AI Providers to configure:

Default Provider — Which integration to use by default for this workspace
Agent Overrides — Route specific agents to specific providers

Workspace Integrations

Integration Types

Provider Resolution

AI Provider Routing

Dashboard Setup

Integration Types

Provider Resolution

Managing Integrations

Create a Workspace Integration

Create a Global Integration

Test a Connection

List Integrations

Secret Storage

AI Provider Routing

Configure Per-Agent Overrides

Resolution Chain

Available Agents

Dashboard

Integrations Settings

AI Providers Settings

What's Next

AI Integration

Git Integration

Embeddings

On this page