Opencloudforce

Opencloudforce is a native Salesforce AI agent system — a flexible, open alternative to Agentforce. It connects your Salesforce org to any LLM available on OpenRouter (Claude, GPT-4, Mistral, Llama, and more), giving your users an intelligent chat interface that can read records, run Flows, and execute Apex — all from within Salesforce.

No external servers. No middleware. Just LWC + Apex + OpenRouter.

---

Benefits

• Multi-model support — switch between Claude, GPT-4, Mistral, Llama, or any other OpenRouter model per agent, without changing code
• Real-time streaming responses — responses stream token-by-token for a fast, natural chat experience
• Tool/function calling — agents can invoke Salesforce Flows or Apex classes as tools, letting them query records, update data, or call external APIs
• Fully native Salesforce — built entirely with LWC and Apex; no external servers or middleware required
• No-code agent configuration — create and configure agents through custom objects; no deployment needed to add new agents or tools
• Conversation memory — configurable conversation history depth keeps context across multiple turns
• Cost-effective — pay only for what you use via OpenRouter, without Agentforce per-user licensing

---

How to Install

1. Create an OpenRouter API Key

Go to openrouter.ai, create an account, and generate an API key from your dashboard.

2. Clone the Repository

git clone https://github.com/your-org/Opencloudforce.git
cd Opencloudforce/package

3. Authorize Your Salesforce Org

sf org login web --alias my-org

4. Deploy to Your Org

sf project deploy start --manifest manifest/package.xml --target-org my-org

This deploys all components defined in manifest/package.xml:

• LWC components: agentHub, agentCard, aiAgentChat, aiChatMessage
• Apex classes: AgentController, AgentToolExecutor, AgentToolCallable, AIQueryRecords
• Custom objects: Agent__c, Agent_Tool__c
• App, tabs, permission set, remote site settings, and CSP trusted sites

5. Set Your API Key

After deployment, update the Custom Label in Setup > Custom Labels:

Label	Value
OpenRouter_API_Key	Your OpenRouter API key

6. Assign the Permission Set

Assign the AI_Agent_User permission set to yourself and any user who needs access to the agent hub:

sf org assign permset --name AI_Agent_User --target-org my-org

Or assign it manually via Setup > Permission Sets > AI_Agent_User > Manage Assignments.

7. (Optional) Create a Demo Query Agent

Run the provided anonymous Apex script to create a realistic demo agent and tool configuration:

sf apex run --file scripts/apex/create_demo_query_agent.apex --target-org my-org

This creates a Sales Ops Assistant (Demo) agent configured to use the query_records tool backed by OCF_AIQueryRecords.

That's it. Open the OCF Agent Studio app from the App Launcher to get started.

---

Objects Reference

Agent__c

Represents an AI agent with its model configuration and behavior settings. Each agent appears as a card in the Agent Hub.

Field	Type	Description
Name	Text	Display name of the agent
OCF_Description__c	Text	Short description shown in the agent selection card
OCF_OpenRouter_Model__c	Text	OpenRouter model ID (e.g. anthropic/claude-3.5-sonnet, mistralai/ministral-8b)
OCF_Temperature__c	Number (0–2)	Controls response creativity. Lower = more deterministic
OCF_Max_Tokens__c	Number	Maximum tokens in each response
OCF_Top_P__c	Number	Nucleus sampling parameter
OCF_Frequency_Penalty__c	Number	Reduces repetition of recently used tokens
OCF_Presence_Penalty__c	Number	Encourages the model to introduce new topics
OCF_Custom_Instructions__c	Long Text	System prompt injected at the start of every conversation
OCF_Enable_Tools__c	Checkbox	When checked, the agent can call configured tools
OCF_Max_Conversation_Turns__c	Number	How many previous messages to include in each API request
OCF_Is_Active__c	Checkbox	Controls whether the agent appears in the hub
OCF_External_ID__c	Text	Optional external system identifier

---

Agent_Tool__c

Represents a callable tool available to an agent. Tools let the LLM take actions in Salesforce — querying data, updating records, running automations, or calling external services. Each tool is linked to a parent Agent__c record.

Field	Type	Description
Name	Text	Display name of the tool
OCF_Tool_Name__c	Text	Technical name the LLM uses when calling the tool (e.g. query_records)
OCF_Description__c	Long Text	Explains to the LLM what the tool does and when to use it
OCF_Action_Type__c	Picklist	Execution method: Flow or Apex
OCF_Action_Name__c	Text	Flow API name or Apex class name to invoke
OCF_Parameters_Schema__c	Long Text	JSON Schema describing the tool's input parameters (used by the LLM)
OCF_Return_Schema__c	Long Text	JSON Schema describing the tool's output (used by the LLM)
OCF_Is_Active__c	Checkbox	Controls whether the tool is available to the agent
OCF_Requires_Confirmation__c	Checkbox	If checked, user confirmation is required before the tool runs
OCF_Timeout_Seconds__c	Number	Maximum execution time for the tool
OCF_External_ID__c	Text	Optional external system identifier

Built-in Tool: AIQueryRecords

Opencloudforce ships with one ready-to-use Apex tool: AIQueryRecords. Set OCF_Action_Type__c = Apex and OCF_Action_Name__c = AIQueryRecords on any tool record to let agents dynamically query any Salesforce object by SOQL.

Accepted parameters:

Parameter	Required	Description
objectType	Yes	Salesforce object API name (e.g. Account)
fields	Yes	List of field names to retrieve
whereClause	No	WHERE condition (without the WHERE keyword)
orderBy	No	ORDER BY clause
limitCount	No	Number of records to return (default 10, max 100)