Configure AI providers and tools

Models

AI providers supply the language or embedding models used by supported Designer capabilities.

Authorisation and packaged capabilities

OAuth apps provide delegated access to external services, while extensions provide prebuilt Greentic tool packages.

Connected and custom tools

MCP servers, component tools and OpenAPI connectors expose external actions to flows and agentic workers. The GitHub App connects repositories to Greentic Actions.

Designer surface
      ↓
Model or tool capability
      ↓
Tenant or team scope
      ↓
Allowed roles, tools or operations
      ↓
Runtime execution

• Open the correct customer tenant before changing AI & Tools settings.
• Confirm whether the capability should be tenant-wide or limited to a team.
• Obtain the required provider credentials, OAuth configuration, server URL, component reference or API specification.
• Decide which Designer surfaces genuinely require access.
• Use test credentials and non-production integrations while validating a new setup.

Enlarge

Enlarge

Type

Select the model capability being connected, such as an LLM or embedding provider.

Provider

Select the model provider. Treat the providers visible in the screenshot as examples rather than a permanent list.

Model

Enter the exact model identifier supported by that provider.

Label

Give the connection a clear administrative name.

Base URL

Leave the default where appropriate, or set the endpoint required by a compatible service, gateway or self-hosted model.

API key

Enter the credential issued for this tenant or environment.

Tenant access

Control whether teams or individual flows may override the tenant default.

Designer roles or surfaces

Grant the provider only to the capabilities that require it.

1. 1Open the tenant.
2. 2Select AI & Tools.
3. 3Open AI providers.
4. 4Select Add provider.
5. 5Choose the provider type and provider.
6. 6Enter the exact model identifier.
7. 7Add a descriptive label.
8. 8Review the base URL.
9. 9Enter the API key.
10. 10Choose the tenant access policy.
11. 11Select the required Designer roles or surfaces.
12. 12Create the provider.
13. 13Confirm that it appears as enabled and has the expected roles.

Enlarge

Provider identifier

The external service being registered.

OAuth application kind

The credential model used by that provider.

Optional team slug

Limits the configuration to a single team when required.

Tenant access policy

Whether teams or flows may override the tenant default.

Client credentials

The client ID and secret issued by the external provider.

Requested scopes

The permissions requested during user authorisation.

Authorisation URL

The endpoint where the user grants access.

Token URL

The endpoint used to exchange and refresh tokens.

1. 1Open AI & Tools → OAuth apps.
2. 2Select Add OAuth app.
3. 3Enter the provider identifier.
4. 4Select the appropriate OAuth application kind.
5. 5Add a team slug only when the configuration should be team-specific.
6. 6Choose the tenant access policy.
7. 7Enter the client credentials.
8. 8Add only the scopes required by the intended tools.
9. 9Confirm the authorisation and token endpoints.
10. 10Create the OAuth app.
11. 11Complete a test user authorisation before relying on it in a solution.

Enlarge

Scope

Selects the tenant default or another supported scope.

Search and filters

Help administrators find an extension by name, publisher, domain or status.

Browse store

Used to discover available extensions.

Setup

Opens extension-specific deployment and credential settings.

Toggle

Controls whether the extension is enabled for the selected scope.

Enlarge

Team

Select tenant-wide or a supported team scope.

Environment

Choose the relevant runtime environment.

Deploy pack ID

Enter the pack identifier expected by the runtime.

Credentials

Add the aliases and secret values required by that extension.

1. 1Open AI & Tools → Extensions.
2. 2Select the correct scope.
3. 3Search for the extension or select Browse store.
4. 4Install the required extension where it is not already installed.
5. 5Select Setup.
6. 6Choose the team and environment.
7. 7Enter and save the deploy pack ID.
8. 8Add the required credentials.
9. 9Close the setup dialog.
10. 10Enable the extension.
11. 11Test one representative tool operation.

Enlarge

Add from store

Select a packaged MCP integration where available.

Add server

Connect a remote MCP endpoint manually.

The two screenshots below are the upper and lower parts of the same Add server form.

Enlarge

Enlarge

Name

A stable administrative name for the server.

Transport URL

The remote MCP endpoint.

Scope

Tenant default or a selected team.

Tenant access

Whether teams or flows may override the tenant default.

Authentication header name

The HTTP header expected by the server.

Authentication token

The sealed credential sent to the server. It may be left blank for a public tokenless server.

Allowed tools

A tool allow-list. An empty value exposes every tool reported by the server.

Enabled

Controls whether the server is available.

Designer surfaces

Select Flow Editor, Agentic Worker or both where supported.

1. 1Open AI & Tools → MCP servers.
2. 2Select Add server.
3. 3Enter a clear server name.
4. 4Enter the transport URL.
5. 5Choose tenant-wide or team scope.
6. 6Review the tenant access policy.
7. 7Add the authentication header and token where required.
8. 8Restrict the allowed tools where possible.
9. 9Enable the server.
10. 10Select the Designer surfaces allowed to use it.
11. 11Create the server.
12. 12Confirm its status and test a permitted tool.

Enlarge

The page has two areas: Component tools, where components are registered, and Registry credentials, where private-registry access is stored.

Enlarge

Name

A clear administrative name for the component.

Source URL

The store, OCI or repository URL of the component.

Component reference

Optional reference identifying the component.

Version

Optional component version.

Digest

Optional digest that pins an exact build.

Allowed operations

An empty value exposes every operation reported by the component, while an explicit list limits what is available.

Enabled status

Controls whether the component tool is available.

1. 1Open AI & Tools → Component tools.
2. 2Add a registry credential first when the source is private.
3. 3Select Add component tool.
4. 4Enter a clear name.
5. 5Enter the store, OCI or repository source URL.
6. 6Add the component reference, version or digest where required.
7. 7Restrict the allowed operations.
8. 8Enable the component.
9. 9Create the component tool.
10. 10Confirm that its operations can be inspected and exposed.

Access a private component registry

Enlarge

Registry host

The host serving the private packages.

Username

The service or bot account used for the pull.

Registry token

A token with only the required package-read permissions.

1. 1Select Add credential.
2. 2Enter the registry host.
3. 3Enter the service or bot username.
4. 4Enter a token with only the required package-read permissions.
5. 5Create the credential.
6. 6Return to the component tool and retry registration.

Registry tokens are write-only secretsOnce saved, registry tokens are not displayed again. Store the original securely and rotate it according to the customer's credential policy.

Enlarge

1. 1Open AI & Tools → OpenAPI connectors.
2. 2Select New connector.
3. 3Enter a concise connector name.
4. 4Select a supported OpenAPI or Swagger specification file.
5. 5Select Upload spec.
6. 6Review the generated connector's status and component information after processing.
7. 7Correct the specification and upload it again if validation fails.
8. 8Register or configure the resulting component as required by the tenant.

Enlarge

Platform administrator

Configures the deployment-level GitHub App credentials.

Partnership administrator

Can connect the tenant only after that prerequisite is complete.

The state shown

The screenshot represents an unavailable prerequisite state, not a successful connection.

1. 1Open AI & Tools → GitHub App.
2. 2Review the displayed status.
3. 3Select refresh to check the deployment configuration again.
4. 4When the page says the GitHub App is not configured on the deployment, contact the Greentic platform administrator.
5. 5Ask the platform administrator to configure the App ID, app slug and private key.
6. 6Return to the tenant and refresh the status.
7. 7Continue with the repository connection controls that become available after the deployment prerequisite is satisfied.

1. 1Add and validate the required AI provider.
2. 2Decide which capabilities are tenant-wide and which are team-specific.
3. 3Configure OAuth or other credentials.
4. 4Install and configure prebuilt extensions.
5. 5Connect MCP servers or register custom components.
6. 6Generate OpenAPI connectors where an existing API specification is available.
7. 7Restrict roles, tools, scopes and operations.
8. 8Test every capability from the Designer surface that will use it.
9. 9Test using a normal tenant user where possible.
10. 10Review secrets and permissions before production use.