Key Concepts

Before diving into configuration, it helps to understand the main concepts that make up the Sprigr Team platform. These building blocks work together to create a flexible AI-powered workspace for your organization.

Agents

An agent is an AI assistant configured for a specific role in your organization. Each agent has a persona (system prompt) that defines its behavior, a set of tools it can use, and access to knowledge bases that inform its responses. Agents are the primary way users interact with AI in Sprigr Team.

Agents are shared across your organization. Any team member with the right permissions can start a conversation with any agent. This makes agents reusable resources rather than personal chatbots.

Every agent maintains context across long conversations and can resume where it left off.

Agent Teams

An Agent Team is a group of agents that can collaborate on tasks. When agents are part of a team, they can delegate work to each other and share conversation context. For example, a general help desk agent might escalate a billing question to a finance agent, which handles it and returns the result.

Agent Teams allow you to build specialized agents for narrow domains while still providing a single entry point for users. The primary agent acts as a router, deciding when to involve other team members based on the conversation.

Companion Agents

A companion agent is a personal AI assistant assigned to each team member. When you sign in to Sprigr Team, your companion agent is automatically created for you. It acts as your personal assistant within the platform - helping you interact with other agents, manage tasks, review workflow approvals, and navigate the system.

Think of it as your PA inside Sprigr Team. Your companion knows your role, your preferences, and your conversation history. It can relay messages to specialist agents on your behalf, summarise information, and route workflow approvals to you when your input is needed.

Knowledge Bases

A knowledge base is a searchable collection of documents and data that agents can reference when answering questions. Knowledge bases use hybrid search - combining traditional keyword matching with semantic vector search - to find the most relevant information.

You can upload documents, sync data from integrations, or push content via the API. Each knowledge base is scoped to your organization and can be attached to one or more agents.

Note

Knowledge bases are powered by the same search infrastructure as Sprigr Search, with support for faceted filtering, custom ranking, and real-time indexing.

Agent memory

Alongside the knowledge bases you manage, each agent keeps its own memory of what it learns in conversation. New memories are written as raw entries and categorised by kind (for example a durable fact about a customer, a preference, or a one-off detail that should expire). Every night Sprigr consolidates that raw memory: durable entries are promoted into the shared company memory, near-duplicates are merged, superseded facts are replaced, and stale entries expire. The result is that an agent recalls the current version of a fact rather than every variant it has ever heard.

Workflows

A workflow is an automated multi-step process that agents can trigger or participate in. Workflows consist of steps (individual actions) and gates that control how a step advances. A gate can run automatically, pause for a human to review or approve, or ask someone to make a decision before the workflow continues.

Workflows are useful for structured business processes like employee onboarding, incident response, content approval pipelines, or customer refund processing. Each step can involve agent actions, API calls, notifications, or human review.

Integrations

An integration connects Sprigr Team to an external service. Integrations are authenticated per-organization using OAuth or API keys and give agents the ability to read and write data in external systems.

Built-in integrations include Gmail, Outlook, Google Calendar, Google Drive, Google Sheets, Google Docs, OneDrive, Microsoft Teams, GitHub, Gorgias, Archibus, AWS, and GCP, with more added regularly. Additional integrations are available as installable Apps from the marketplace. Each integration exposes a set of tools that agents can use during conversations.

Channels

A channel is a communication endpoint where users can interact with agents. Sprigr Team supports several channel types:

• WebChat - The built-in chat surface in the Sprigr Team portal. Available on every agent with no configuration.
• Telegram - Bot-based interaction for Telegram users.
• WhatsApp - Messaging through the platform-managed Sprigr number, available for companion agents.
• Slack - Connect a workspace from the agent’s Slack channel card with the Add to Slack button, or install the Slack marketplace app. Both routes work.
• Website chat - A chat widget on a website hosted by Sprigr, so visitors can talk to an agent.
• Webhook - Inbound webhooks that deliver a payload to an agent or workflow.
• MCP - External AI clients (Claude Code, Claude Desktop, Cursor) talking to an agent through the MCP server.
• Internal - Agent-to-agent traffic such as delegation, which never leaves the platform.
• Voice chat - Spoken replies and a hands-free mode inside the portal chat, see Chat.

Customer conversations from these channels land in the Inbox in the Sprigr Team portal, so your team can monitor and manage them from one place regardless of where they originated. Conversations you have directly with your own agents live under Chat.

Webhooks

Inbound webhooks allow external systems to trigger agent actions or workflows by sending HTTP requests to a unique webhook URL. Each webhook can be configured with a custom script that processes the incoming payload and determines what action to take.

This is useful for event-driven automation - for example, triggering an agent response when a new support ticket is created in an external system, or kicking off a workflow when a deployment completes.

MCP (Model Context Protocol)

Sprigr Team supports the Model Context Protocol (MCP), exposing your entire platform as an MCP server with over 100 tools that external AI clients can use. Clients like Claude Code, Claude Desktop, Cursor, or other MCP-compatible tools can connect to your workspace and manage agents, workflows, knowledge bases, marketplace apps, schedules, and more - all without opening the portal.

Most MCP tools call the platform API directly with zero agent token cost. send_message is the one that routes through an agent for an AI-powered reply; poll_message (which fetches that reply) and notify_chat (which posts into team chat) call the platform directly like the rest. Each API key is linked to the user who created it, inheriting their role-based data visibility - a member’s key only sees their assigned agents, while an admin’s key sees everything.

Each organisation can generate multiple API keys with different scopes (read, write, or both), optionally restricted to specific agents.

Display size and timezones

The portal has a one-click display size control in the top bar that cycles between compact (the default), cozy, and comfortable. Your profile shows the timezone detected from your browser, and schedules, task history, and other timestamps across the portal are rendered in your organisation’s timezone.