Core concepts
RedFlare brings your conversations, agents, and automations into one place. This page is a quick orientation to the handful of ideas you'll meet everywhere in the app, so the rest of the docs feel familiar from the start.
Keep this page open as a glossary while you explore. Each concept maps to a section in the left sidebar, grouped under Explore, Build, and Connect.
What you can do
- Upload and review Conversations, then process them with a Playbook to get summaries, scores, and severity.
- Tell apart your Agents (the people on the calls) from AI Agents (the assistants you configure to handle tasks).
- Build no-code Workflows that start from a trigger and run a series of steps.
- Organize reference material into Knowledge collections and articles.
- Connect external services through Integrations, Webhooks, and Phone numbers.
- Give your AI Agents extra abilities with System tools, and stay informed with Alerts.
How it works
Underneath the friendly sidebar, RedFlare runs a connected pipeline that turns raw recordings into structured insight, then lets automations act on it.
- Conversations come in from a connected integration, an uploaded recording, or a phone number. Each one is an interaction that moves through clear states: Pending, Processing, Processed, and (when something goes wrong) Retrying or Failed.
- The audio is transcribed. RedFlare converts speech to text and separates who said what, labelling each turn as the agent, the customer, or an automated IVR system. It captures direction too (Inbound or Outbound) and details like audio duration.
- A Playbook scores the call. The playbook you choose runs its checks against the transcript and audio, detecting events, applying rubrics, and producing a Quality score, a Severity (Critical, High, Medium, or Low), and a Summary. Some checks are AI-driven and some are rule-based, and each run is saved as a Playbook version so you can compare results.
- Enrichment adds context such as sentiment, translation for non-English calls, and language analysis, so the same conversation reads consistently no matter where it came from.
- Automations react. Once a conversation finishes processing, Workflows and Alerts can pick it up, filter it, process it further, send notifications, or export a report.
Because every step records its status, you always see where a conversation is, and RedFlare can automatically retry a step that fails rather than losing the call.
The building blocks
Organization and workspace
When you sign in, you're working inside your organization's workspace. Everything you create, including conversations, workflows, agents, and knowledge, lives here and is shared with your team. The sidebar is your home base: it's split into Explore (review what's happening), Build (create automations and assistants), and Connect (wire up outside services).
Conversations
A conversation is a call or interaction that RedFlare has captured or that you've brought in yourself. Open Conversations from the sidebar to see them all in a table you can sort, filter, group, and save as named Views.
Each conversation carries useful detail once processed, including a Severity, a Quality score, an Agent, the Audio duration, a Transcription, and a Summary. The transcript is split by speaker, so you can follow the back-and-forth between the agent and the customer. To bring in your own recordings, use Upload Calls.
Playbooks
A playbook tells RedFlare how to process a conversation. When you upload or reprocess calls, you'll be prompted to "Select a Playbook" with the note: "Selected playbook will be used to process the uploaded calls."
Think of a playbook as your processing recipe. It's built from reusable pieces, the checks (events and features) it looks for, the rubrics it scores against, and the summary template it fills in. Different playbooks produce different scores, summaries, and insights, and you can switch between Playbook versions to compare results. As the app explains: "Each version represents a run of the call through a playbook configuration. Switch versions to view results from different processing runs."
Agents vs. AI Agents
These two sound alike but mean very different things.
- Agents are the people handling conversations. Open Agents to "View and manage your complete list of active agents and their statuses," along with stats like Total conversations, Average conversation duration, and User satisfaction rating.
- AI Agents are intelligent assistants you build and tune. As the app puts it, you "Configure and tune intelligent agents to handle tasks and automate workflows." You can give each one System tools to expand "What this agent can do," and point it at a Knowledge base for grounding.
Workflows
A workflow is an automation you assemble visually. Every workflow starts with a Trigger ("Select a trigger to start a workflow"), such as an integration sync starting, completing, or failing, or a Schedule. From there you chain together steps using Add new step, choosing from blocks like Select conversations, Filter, Sort by, Process conversations, Send alert, and Report export.
When you're happy with it, choose Publish workflow to make it "live and available for execution," or Pause workflow to take it offline without losing your setup. Each published version is tracked, and every run records its progress step by step, so you can see exactly what ran and where anything stopped.
Knowledge
The Knowledge Base is where you "Turn your data into intelligence." It's organized into two levels:
- Collections group related material. You "Group and organize multiple knowledge bases into a single collection."
- Articles are the individual entries inside a collection. You can upload documents or connect URLs, and keep a Version History for each one.
Once added, RedFlare processes each article so its content can be searched and used by your AI Agents. If a version can't be processed, it's clearly marked: "This version failed to process and cannot be viewed."
Integrations
Integrations connect RedFlare to the tools you already use. As the app describes it, you "Connect agents to external services to sync data and automate workflows across platforms." Integrations cover contact centres, customer relationship tools, and other data sources.
Each connection shows its Last synced time and Next sync in, and you control how it runs with a Sync Mode that "Determines how data should synced." Choose Append to add new records without changing existing ones, or Full Refresh to replace everything. You can schedule syncs (manual, basic, or by a cron expression), trigger one yourself with Sync now, and review every run in Sync History.
Take a quick tour
- Select Conversations in the sidebar to see your interactions.
- Open one to read its Summary, Transcription, and scores.
- Switch to Workflows under Build to see how automations are assembled.
- Visit AI Agents to see the assistants configured for your workspace.
Upload your first conversation
- In the sidebar, hover Conversations and choose Upload Calls.
- Add your recording. RedFlare validates the file and its details before uploading.
- When prompted, choose Select a Playbook, the playbook decides how your call is processed.
- Confirm to start processing. The conversation appears as Processing and switches to Processed once its transcript, scores, and summary are ready.
Reprocess a conversation
- In Conversations, select one or more processed calls.
- Choose Reprocess conversation.
- Pick the playbook to run. RedFlare notes that you've "Selected conversation(s) to process."
- Confirm. The new run is saved as a fresh Playbook version alongside the previous one, so you can compare results.
You can process unprocessed calls or reprocess already-processed ones, but not both in the same action: "Choose either unprocessed calls or processed calls to reprocess, not both."
Create a Collection in Knowledge
- Select Knowledge in the sidebar.
- Choose New Collection (or Create New Collection from the empty state).
- Enter a collection name and description.
- Open the collection and add a New Article by uploading a document or entering a URL.
Connect an integration
- Select Integrations under Connect.
- Choose New Integration.
- Pick the service you want to link and follow the prompts.
- Set a Sync Mode and schedule, then check Last synced to confirm data is flowing.
Tips
Save your most-used conversation filters as a View so you can jump straight back to them. RedFlare can even Autosave views as you work.
Remember the difference: Agents are the people on the calls, while AI Agents are the assistants you configure. Whenever a screen talks about handling tasks or automating work, it means AI Agents.
A workflow only runs once you Publish it. If results stop appearing, check whether the workflow is Paused, and open a run to see which step stopped.
A playbook drives how every conversation is scored and summarized. If two conversations look different, check whether they were run through different playbooks or Playbook versions.
If a conversation shows as Failed or Retrying, RedFlare attempts the processing step again on its own. You can also reprocess it manually once the underlying issue (such as a missing or unreadable recording) is resolved.