Relaticle: The First Open-Source CRM Built Agent-First from the Ground Up

In the CRM space, most products simply bolt AI features onto existing systems. Relaticle takes a fundamentally different approach — it was designed from the ground up for AI agents, making it a truly agent-first open-source CRM that has earned 70 upvotes on Product Hunt.

Agent-First Architecture

Relaticle's most distinctive feature is its "agent-first" design philosophy. The system exposes 37 native tools to any MCP (Model Context Protocol) client via OAuth. These tools automatically detect custom fields defined in your workspace and incorporate them into the agent's schema.

The benefits are immediately practical: developers can connect to Relaticle using the standardized MCP protocol without building custom integrations for each AI application. Any custom fields defined in a workspace are automatically synced into the agent's context, dramatically lowering the barrier to building AI-powered applications.

MCP (Model Context Protocol) is an open standard proposed and open-sourced by Anthropic in late 2024, designed to solve the interoperability problem between large language models and external tools and data sources. Think of it as the "USB port" of the AI world — whether you're using Claude, GPT, or any other compatible model, you can call external systems through a unified MCP interface without writing separate adapters for each model. MCP uses a client-server architecture: the MCP Server exposes tools and data, while the MCP Client (typically an AI application) discovers and invokes them. Relaticle implements itself as an MCP Server, meaning any MCP-compatible AI client — such as Claude Desktop or Cursor — can directly interact with CRM data like contacts, deals, and tasks without any glue code.

Approval Gating: A Safety Mechanism for AI Write Operations

Unlike fully autonomous AI operations, Relaticle enforces stricter controls within its in-app assistant. Every write operation initiated by an AI generates a "proposal" that clearly displays the exact changes about to be made, requiring the user to approve each one individually.

This mechanism strikes a balance between AI automation and human oversight. It preserves the efficiency gains that AI provides while using a human approval step to prevent potential errors or unintended modifications. For enterprise CRM use cases, this approval gating delivers the data security guarantees that organizations require.

Fully Open Source with Flexible Deployment Options

Relaticle is licensed under AGPL and supports fully self-hosted deployment — even local inference via Ollama, with no dependency on cloud-based AI services. This is significant for data-sensitive organizations: customer data can remain entirely within the enterprise without being uploaded to any third-party cloud.

On the commercial side, Relaticle offers two options:

• Self-hosted: Free forever
• Cloud-hosted: Priced per workspace, not per user seat

This pricing model is more attractive for larger teams, eliminating the problem of costs scaling linearly as headcount grows.

Ollama is a tool that lets users run open-source large language models locally — including Llama, Mistral, Qwen, and other leading open models — without requiring an internet connection for inference. In a CRM context, pairing Ollama with Relaticle means organizations can build a completely offline AI pipeline: customer data stays in a local database, AI inference runs on a local GPU or CPU, and nothing passes through any cloud node. AGPL (GNU Affero General Public License) is a "strong copyleft" open-source license that requires any derivative product providing network services based on the code to also be open-sourced. This protects the rights of open-source authors, but it also means that companies wishing to commercially fork Relaticle and offer it as a SaaS product must comply with the same open-source requirements.

Native MCP Support and Technical Innovation

Relaticle's emergence represents a rethinking of how the CRM industry approaches AI integration. Traditional CRM products typically build a complete system first, then figure out how to "add" AI later. Relaticle designs the entire system architecture around how AI agents work.

Native MCP support gives it strong interoperability. The 37 tools cover core CRM operation scenarios, and combined with automatic schema synchronization, developers can rapidly build customized AI workflows. The approval gating mechanism maintains automation efficiency while satisfying enterprise requirements for data operation security.

For the open-source community, a reference implementation of an agent-first CRM holds significant value. Developers can draw on Relaticle's architectural design to explore more possibilities for deeply integrating AI with business systems.

Who It's Built For

Relaticle is particularly well-suited for:

• Technical teams that need deep CRM customization
• Organizations with strict requirements around customer data security
• Developers looking to explore real-world AI agent deployments in business systems
• Small and mid-sized teams seeking predictable, controllable costs

As a rising product on Product Hunt, Relaticle has received positive feedback from the community. As the MCP protocol ecosystem matures and AI agent adoption continues to accelerate, business systems designed from the ground up for AI may well become the new direction for the industry.