OpenTag

Product Introduction

1. Definition: OpenTag is a model-agnostic AI agent and collaboration platform designed for team chat environments like Slack and Microsoft Teams. Technically, it is a cloud-based orchestration layer that routes user requests to the most suitable AI model from a pool of over eighty options, executes actions via connected tools, and automates workflows.
2. Core Value Proposition: It exists to function as a proactive AI coworker that reduces operational toil and knowledge loss. Its primary value is in automating repetitive tasks, maintaining a living company wiki, and intelligently routing work to cost-efficient AI models, thereby cutting AI operational costs by up to 70% compared to using only frontier models like Claude or GPT.

Main Features

1. Cloud Agents in Team Chat: OpenTag deploys ephemeral, sandboxed cloud agents that operate directly within Slack or Microsoft Teams channels. How it works: When tagged, it spins up a secure, scoped execution environment with access only to pre-connected tools (like HubSpot, Stripe, GitHub). The agent performs the task, reports progress in-thread, and is torn down upon completion, ensuring security and context isolation.
2. Proactive Automation & Scheduling: The platform learns from team behavior to suggest automations. How it works: It uses pattern recognition to identify repetitive manual requests (e.g., "pull Monday numbers"). After observing the same request multiple times, it proactively offers to own and schedule the task (e.g., "Run this every Monday at 9?"). This transforms one-off commands into managed, recurring workflows without manual configuration.
3. Self-Organizing Company Wiki: OpenTag automatically generates and maintains documentation. How it works: It continuously parses decision-making conversations across connected channels. It synthesizes this context into structured wiki pages (e.g., refund-policy.md). When a policy changes in a discussion, it revises the corresponding page, archives the old version, and notifies relevant team members, ensuring documentation is always current and traceable.
4. Model-Agnostic AI Routing (via Conifer): This is the core technical differentiator. How it works: Instead of being locked to a single provider (e.g., Anthropic's Claude), OpenTag uses Conifer's model catalogue to dynamically route each task. It evaluates factors like task complexity, required capability, cost, and latency to select from over 80 models, including Claude Opus, GPT-5.6, Gemini 3.1 Pro, DeepSeek, Llama 4, and Mistral. This ensures optimal performance and cost-efficiency per task.

Problems Solved

1. Pain Point: High and unpredictable costs from using premium frontier AI models (Claude, GPT-4) for all tasks, including simple ones. Related Keywords: AI agent cost, model spend optimization, Claude Tag alternative.
2. Pain Point: Knowledge silos and outdated documentation, where critical company processes and decisions are buried in chat history, leading to repeated questions and information loss during employee turnover.
3. Pain Point: Manual, repetitive work that interrupts deep focus, such as pulling weekly reports, chasing invoices, or updating dashboards, which creates operational drag.
4. Target Audience: Operations Managers, Engineering Team Leads, Sales Operations Specialists, Customer Support Managers, and startup founders who need to scale processes without linearly scaling headcount.
5. Use Cases:
   • Automated Reporting: Automatically generating and posting weekly sales or growth metrics to a designated Slack channel every Monday.
   • Customer Workflow Automation: Processing refunds by checking policies, pulling Stripe data, and creating tickets in Zendesk upon approval.
   • Onboarding & Knowledge Retrieval: New hires can ask OpenTag in a channel for the current refund policy or deployment runbook, receiving an answer sourced from the auto-updated wiki.
   • Proactive Issue Investigation: Monitoring error channels and automatically pulling relevant logs from PostHog or GitHub to pre-investigate issues before human intervention.

Unique Advantages

1. Differentiation vs. Claude Tag / GPT-based Assistants: Unlike vendor-locked tags (Claude Tag), OpenTag is model-agnostic. Its business model is not tied to selling model tokens, aligning its incentives with cost-effective task completion for the user. It also offers multi-platform support (Slack & Teams) and deeper workflow automation compared to basic Q&A assistants.
2. Differentiation vs. Traditional RPA/No-Code Tools: OpenTag requires no manual workflow building. It learns by observation and suggests automations, operates conversationally within existing collaboration tools, and integrates AI decision-making into the execution steps.
3. Key Innovation: The combination of context-aware model routing and proactive automation discovery. The system doesn't just execute a command; it analyzes the task's requirements to choose the most economically and technically suitable AI model from a vast fleet, and it observes team patterns to automate work before being explicitly asked.

Frequently Asked Questions (FAQ)

1. How does OpenTag save 70% on AI model costs? OpenTag saves costs by intelligently routing tasks away from expensive frontier models (like Claude Opus) for simple work. It uses a vast catalogue of specialized and more efficient open-source models (via Conifer) for tasks they can handle competently, reserving premium models only for complex reasoning jobs, leading to significant average cost reduction per task.
2. Is OpenTag secure for connecting to tools like Stripe and databases? Yes, OpenTag employs a security model where each agent run is sandboxed and scoped. Permissions follow the user who made the request, and agents only have access to the tools explicitly connected for that workspace. Sensitive data is not retained after the task is complete.
3. Can OpenTag work with our private/internal AI models? The platform's architecture, through its integration with Conifer's model router, is designed to access a broad catalogue. While the current offering focuses on the listed 80+ public/cloud models, the model-agnostic foundation suggests potential for future support for routing to private model endpoints.
4. What happens if the chosen AI model provides a poor-quality result for a task? OpenTag's routing logic is based on reliability and capability metrics. For critical tasks, users can provide feedback. The system's learning and routing algorithms are designed to optimize for successful task completion, potentially re-routing similar future tasks to more capable models based on performance history.
5. How does the self-updating wiki handle conflicting information from different channels? The wiki system sources information from conversations and likely uses semantic analysis to identify the most recent and authoritative statements on a topic. When conflicts arise, it may prioritize information from designated source channels or flag the conflict for human review, maintaining a revision history for auditability.