Find answers before opening a ticket

Your self-service operating manual for agentic business work.

Learn how to build agents, upload knowledgebases, configure semantic retrieval, manage tools, and keep AI workflows governed without waiting on support.

Build your first agent

Create an agent, write its system prompt, choose tools, save it, then run a preview before adding it to workflows.

Open Agent Builder

Choose tools

Add memory

Run preview

Create knowledge memory

Paste policies, operating notes, customer context, or URLs. Beam prepares the content for semantic retrieval.

Select an agent

Add source text

Build knowledge index

Preview retrieval

Upload documents

Upload text-compatible files. Beam stores the source, extracts content, and makes it searchable for the selected agent.

Choose a file

Upload source

Prepare semantic chunks

Review status

Use governed tools

Tools are selected per agent and validated server-side, so workflows only expose approved actions.

Opportunity score

Workflow blueprint

Policy risk

ROI projection

Semantic Knowledge

Search index configuration

Create this semantic search index named knowledge_chunks_vector_index on the knowledge chunks collection.

{
  "fields": [
    {
      "type": "vector",
      "path": "embedding",
      "numDimensions": 1536,
      "similarity": "cosine"
    },
    {
      "type": "filter",
      "path": "userId"
    },
    {
      "type": "filter",
      "path": "agentId"
    },
    {
      "type": "filter",
      "path": "documentId"
    }
  ]
}

FAQ

Self-service answers

Where do uploaded knowledge files live?

Original files are stored securely. Extracted text, chunks, metadata, and semantic representations are scoped by user and agent.

What semantic index do I need?

Create the semantic knowledge index using the configuration shown below. The app also falls back gracefully in local development if the index is not present.

How does an agent use knowledge?

When an agent preview or workflow run starts, Beam retrieves relevant knowledge for the selected agent and prepends the best matches to the model context.

Can I disable a risky agent?

Yes. Set the agent status to Disabled in Agent Builder. Disabled agents cannot be used for live generation.

Preparing workspace

{ "fields": [ { "type": "vector", "path": "embedding", "numDimensions": 1536, "similarity": "cosine" }, { "type": "filter", "path": "userId" }, { "type": "filter", "path": "agentId" }, { "type": "filter", "path": "documentId" } ] }