Redbird

Redbird's connectors to Salesforce, HubSpot, Pipedrive, Dynamics 365, Intercom, Snowflake, BigQuery, Databricks, Slack, Marketo, Zendesk, Jira and many more support write back as well as reads, updating records, creating entries such as HubSpot follow up tasks and triggering actions, with 60 plus integration guides in the docs.

Redbird also connects to Redshift and Azure Synapse, PostgreSQL, MySQL, SQL Server, MongoDB and Oracle, Amazon S3, Google Drive, SharePoint, Azure Blob and Dropbox, and Google Analytics, and can connect to any system with a documented REST, GraphQL or SOAP API. Results can be pushed to cloud storage or a data warehouse, sent by email, or exported to PowerPoint.

SourceRedbird, redbird.io/product/connectors and docs.redbird.io llms.txt indexread 2026-10-05

Work runs as workflow canvases of nodes that the team builds, views and edits, with AI agents placed as standalone nodes with defined inputs and outputs, so several agents and steps chain in one flow that the team sets up.

Autopilot builds a workflow from a chat request, brings in specialist agents when their invoke conditions match, and runs all of a workflow's nodes or only the ones a user picks.

Signals announce success, failure, data updates or custom conditions, and checkpoints hold a workflow until signals combine with and, or and not logic within an optional time window, so a report rebuild can be skipped when the data never arrived.

SourceRedbird, docs.redbird.io Build and run workflows, Add advanced workflow logic and Talk to Autopilotread 2026-10-05

An admin layer lets teams load data ontologies (dataset descriptions, field definitions, relationships), business logic for calculations and reporting blueprints that the agents work from. Data lands in Redbird datasets and file collections that agents query, and published chat environments add an AI Context resource for reference material beside the shared outputs. Specialist agents get read only data tools automatically.

SourceRedbird, redbird.io platform post and docs.redbird.io Datasets, Publish a chat environment and Add specialist AI agentsread 2026-10-05

Teams can view and modify what the AI builds, editing any step by point and click or code, and role based access limits who can do what. No step requires a person to approve before an agent's write back or replayed action runs. In Autopilot a preview panel shows the work in progress, and a user can select a chart, slide or dataset and ask for a specific change before it is used. Builders who publish a chat environment choose which outputs end users see, while the connections, preparation steps and schedules behind them stay private.

SourceRedbird, redbird.io and docs.redbird.io Talk to Autopilot and Publish a chat environmentread 2026-10-05

Redbird states it is SOC 2 Type II certified and includes SSO and SAML, role based access controls and full audit logging out of the box, along with two factor authentication and node level access. Each user can set up two factor authentication, and Redbird says it never trains models on customer data.

SourceRedbird, redbird.io and docs.redbird.io Two-Factor Authenticationread 2026-10-05

Redbird keeps a full audit trail for every agent action, with each step logged, explained and inspectable in the workflow it built, and the API reports each workflow run's status. Runs are listed in a Workflow Runs panel, and each run executes every node in order, pulling fresh data, reapplying the preparation steps and updating outputs such as reports.

SourceRedbird, redbird.io and docs.redbird.io Build and run workflowsread 2026-10-05

AI Chat works through dynamic prompts within a conversation, and the workflows agents build are saved and rerun deterministically, but the agents keep no memory across sessions. Custom specialist definitions, with their rules and conditions for when to step in, stay in place until a team switches them off or deletes them.

SourceRedbird, docs.redbird.io Add specialist AI agents and Build and run workflowsread 2026-10-05

The platform can run cloud hosted, on premises or in a private VPC. No region list for the hosted service is published.

SourceRedbird, redbird.ioread 2026-10-05

Ready made specialist agents include a SQL agent, an unstructured data agent, a data visualization agent, a fuzzy matching agent, a text autotagger and a multimedia autotagger. Redbird Specialists are prebuilt experts, such as the Data Science Specialist for forecasting and modeling, that Autopilot calls in when a request matches, beside custom specialists a team defines with a name, a purpose, a description of when to invoke it and its own logic. An AI Autotagging operation and data science models also ship as workflow steps.

SourceRedbird, docs.redbird.io Add specialist AI agents and llms.txt indexread 2026-10-05

Whole workflow canvases run on automated runs, which the API describes as scheduled execution configurations, and an outside system can start a canvas run or a single node with an authenticated API call. A schedule can be set in plain language, such as every Monday at 9am, and edited in the Automated Runs panel. Signals fire on success, failure or a data update, and Email Collect brings in data from forwarded email.

SourceRedbird, docs.redbird.io API Tutorial, Build and run workflows and Add advanced workflow logicread 2026-10-05

A proprietary domain specific language handles analytics steps. Redbird names no model provider and offers no model choice or routing that a customer can set. Autopilot and specialist agents have no model setting, and Redbird says it never trains models on customer data.

SourceRedbird, redbird.io and docs.redbird.io Talk to Autopilot and Add specialist AI agentsread 2026-10-05

A REST API at app.redbird.io/api uses token authentication and exposes endpoints to run workflow nodes, query datasets and write rows into collections, with an SDK authentication page in the docs. Builders can also publish a chat environment and share it, giving end users a self serve chat over chosen outputs with the same specialist agents.

SourceRedbird, docs.redbird.io API Tutorial, Publish a chat environment and llms.txt indexread 2026-10-05

Self healing agents repair steps at run time when data, APIs or interfaces change. Redbird publishes no test set, no scoring of agent output and no check that must pass before a change goes live. Checkpoints can hold everything downstream when the numbers look wrong, so a report is not rebuilt on bad data.

SourceRedbird, redbird.io and docs.redbird.io Add advanced workflow logicread 2026-10-05

Through a browser extension a person records a task done in a web interface and the agent replays it on its own, adapting to minor interface changes, for data extraction and data entry in systems without an API. The agent analyzes the recording and learns the intent behind each step before it replays the task.

SourceRedbird, redbird.io/product/connectorsread 2026-10-05