Maple

Layer 02 · Reasoning + Agents

Reasoning + Agents.Where decisions and actions happen.

Claude is the reasoning engine. Agentforce and Maven AGI are the runtimes that take action: qualifying leads, resolving cases, orchestrating revenue. Two deployment paths, one quality bar.
Book a Strategy SessionSee the full stack

The agent layer is where intent becomes action. A reasoning engine (Claude) decides what should happen; a runtime (Agentforce or Maven AGI) makes it happen — touching records, sending messages, escalating cases, opening tickets, closing deals.

Maple deploys this layer in one of two configurations depending on your trust boundary, existing investments, and the work the agent has to do. Native: Claude embedded directly inside Agentforce via the Salesforce-Anthropic integration — your data never leaves your Salesforce trust boundary. Standalone: Claude on AWS Bedrock or direct API for stacks where Agentforce isn't the right fit. Both are first-class.

Anthropic ClaudeSalesforce AgentforceMaven AGI

What We Deliver

Specifically, in an engagement.

  • Native path: Claude inside Agentforce via the Salesforce-Anthropic integration: data stays inside your Salesforce trust boundary, no external API hop
  • Standalone path: Claude on AWS Bedrock or direct API for non-Salesforce stacks: same Claude reasoning quality, different runtime
  • Maven AGI for L1 support deflection: alternative to Agentforce on customer-service-heavy use cases, or complement for high-volume tier-1
  • Industry-tuned reasoning patterns (FinTech compliance prompts, HealthTech triage flows, B2B SaaS lifecycle agents)
  • Tool-use orchestration: agents that call Salesforce APIs, MCP servers, and your internal systems with proper auth + audit logging
  • Agent eval harness: test cases, regression suite, and human-in-the-loop review patterns shipped with every deployment

How it Changes by Industry

Same architecture. Different tuning.

B2B SaaS

Lifecycle agents that qualify pipeline, draft renewal motions, and deflect support: Claude + Agentforce native, with Maven AGI for tier-1.

FinTech

Claude on AWS Bedrock inside your VPC, or native inside Agentforce. Either path keeps regulated data inside the trust boundary your compliance team already approved.

HealthTech

Triage agents, patient outreach orchestration, and prior-auth automation. HIPAA-compliant by architecture, not by retrofit.

How It Connects

The other layers in the stack.

Data Foundation

The unified data layer the agents reason on. Without it, the agent layer is just an expensive chatbot.

Explore

Experience Layer

Where the agent's work surfaces: voice, web, embedded UI, Slack, WhatsApp.

Explore

Agentic Governance

Recurring tuning, eval, and audit so agents don't quietly degrade in year one.

Explore

Ready to architect your agentic stack?

Take the Readiness AssessmentBook a Strategy Session