AI Agents
Provide your AI agent with a goal, the context it needs and access to the right tools so that instead of scripting every decision, the agent can determine what to do next while staying inside the workflow you define.
Build agents with the right context and controls
Agents can use models, memory, knowledge retrieval, tools and guardrails to handle more complex processes while keeping each part configurable.
Question
Pause a run and ask for input instead of guessing.
Task
Delegate a complex task to a specialized sub-agent.
Auto Memory
Keep useful facts available across conversations.
Skills
Give agents reusable knowledge and capabilities.
Give AI enough freedom to be useful without giving up control. Keep people involved when decisions matter, ground responses in your own knowledge and let agents take on more complex goals when you're ready.
Keep humans in the loop
Human-in-the-loop steps let you pause a workflow, ask for approval or collect missing information before execution continues.
AI Agents can also use these same steps as tools, so an agent can ask for human confirmation before taking an action that needs a person in the loop.
Approve an action
Pause before taking sensitive actions.
Request information
Ask for missing information.
Ground AI in your knowledge
Connect your documents and make relevant business knowledge available to workflows and AI Agents when they need it.
Instead of relying on the model's knowledge alone, agents can retrieve relevant information from your own content.
Combine AI Agents with flow controls to build common agentic patterns yourself today, from simple chains to parallel and orchestrated multi-agent workflows using patterns like:
Test your agent before you trust it
Build and test your agent side by side. Try real prompts, see which tools it calls and inspect the inputs and outputs before putting the agent into a production workflow.
How AI Agents work in ByteChef
Explore how ByteChef combines models, tools, memory, knowledge and guardrails to build AI-powered workflows.
Bring Your Own LLM
Use the AI providers and models you already trust. Connect providers such as OpenAI, Anthropic, Azure or Mistral and choose the model that fits each workflow.
Configure AI providers your way
A self-service AI Providers catalog under Settings lets admins activate providers and manage their credentials for each environment, in every edition. Providers can also be configured through environment variables and configuration properties.
Use AI across your workflows
Use different models for different tasks without rebuilding your workflows. ByteChef's universal AI components support text and image use cases across supported providers.
AI Security & Governance
AI should be powerful without becoming a blind spot. ByteChef lets you place guardrails around AI and check incoming requests and outgoing responses before they continue.
Configure guardrails for your use case - block harmful prompts before they reach the AI and redact sensitive data before it reaches your users.
One policy for the whole workspace
Instead of configuring guardrails agent by agent, set one workspace-level content policy that applies to every LLM call the instance makes: redaction, blocked terms, moderation and injection detection, on the way out and on the way back.
Fail safely, not silently
Guardrails are designed to fail closed: if a safety check cannot run, the request is blocked rather than passed through without protection.
See what was blocked
Blocked requests and guardrail failures are logged so your team can understand what happened and investigate when needed.
Human in the Loop
Not every decision should be fully automated. ByteChef allows workflows to pause and wait for human review or approval before continuing execution - routing the request to Slack, email or another channel and resuming automatically once a decision is made.
This gives organizations the flexibility to automate routine work while keeping people involved in high-value, high-impact or sensitive business decisions.
Financial approvals
Route spend or budget decisions for approval before they continue.
Content review
Review AI-generated content before it is sent or published.
Customer onboarding
Pause onboarding workflows when manual verification is required.
Sensitive updates
Require human review before sensitive records are changed.
Approval steps can be placed where a workflow needs a human decision, allowing reviewers to approve, reject or provide information before execution continues.
Start with an AI workflow
Explore ready-to-use AI workflows and adapt them to your own processes instead of starting from scratch.
Classify Gmail emails with OpenAI and forward them to the right team
Polls Gmail for new emails and uses OpenAI to classify each one into a single department label (Sales, Support, Finance, Operations, HR, or URGENT). The matching Gmail label is applied, the department's recipient address is stored, and the original email is forwarded to the correct team.
Score CVs from Google Forms with OpenAI and sort them in Google Drive
When a Google Forms response is submitted, downloads the attached CV PDF from Google Drive, extracts its text, and summarizes it with AI. OpenAI scores the candidate from 0 to 10 against a Senior Java Developer job description. The name, email, summary, and score are recorded in a Google Sheet, and the CV file is filed into a Rejected or Potential candidates Google Drive folder based on the score.
Load a folder of Markdown files into Pinecone for RAG
Lists all files in a local documentation directory recursively, then loops over each file, reads its contents, and loads it into a Pinecone vector database as a Markdown document, building a searchable knowledge base for retrieval-augmented generation.
Don't miss what comes next
Curious what's coming or waiting for a feature to ship? Subscribe to be the first to know when it launches, plus get release notes, new feature announcements and subscriber-only discounts.
By subscribing, you agree to receive emails from ByteChef and to our Privacy Policy. You can unsubscribe at any time.
