HomeCaveauAIAgent Builder
CaveauAI Feature

Agent Builder: chain your AI tools into automated pipelines

Define multi-step workflows that search your corpus, analyze results, and deliver findings by email or webhook — on a schedule or on demand.

Cron Scheduling 5 Step Types Email + Webhook Delivery

Three ready-made pipelines. Customize from there.

Each template chains CaveauAI tools into a multi-step workflow. Pick one, adjust the steps, set a schedule, and let it run.

Research Assistant

Search your corpus with configurable depth (quick, standard, deep), analyze findings with source citations, synthesize a structured digest, and email it to your team.

search chat transform email
3 depth modes 2–5 steps

Compliance Monitor

Monitor multiple regulatory topics across your document corpus, run analysis on each finding, and send email alerts when something surfaces. Built for legal, finance, and governance teams.

search chat transform email
Multi-topic 4 steps

Document Processor

Search documents by topic, analyze them with full citations and structured summaries, and deliver output via email or webhook. Good for report generation and automated briefings.

search chat email / webhook
Citation mode 2–4 steps

Define, schedule, execute, deliver

Every agent follows the same lifecycle: configure the pipeline, set a trigger, let the engine run, and get the results where you need them.

Define
Template or custom
Configuration
Schedule
Cron or manual
Timezone-Aware
Execute
Up to 3 concurrent
Pipeline Engine
Deliver
Email or webhook
Notification

Five building blocks for any workflow

Combine these steps in any order. Each step can reference the output of the ones before it using template variables.

01

Search

Query your CaveauAI corpus using semantic vector search. Returns ranked document passages with source citations and relevance scores. Supports single queries, multi-query iteration, and category filtering.

Corpus Search
02

Chat

Send a prompt to the AI model with context from previous steps. Use template variables like {{steps.search.output}} to inject prior results into the prompt.

LLM Inference
03

Transform

Reshape, filter, or format the output of previous steps locally without an API call. Create structured summaries, extract key findings, or prepare data for email delivery using Jinja2 templates.

Data Transform
04

Email Notification

Send the pipeline results as a formatted HTML email digest styled in the CaveauAI design language. Supports subject line templating and multi-recipient delivery.

Email Delivery
05

Webhook

POST the pipeline output to any public URL as a JSON payload. Integrate with Slack, Microsoft Teams, ticketing systems, or internal APIs. Includes SSRF protection blocking private IP ranges.

Webhook POST

Three workflows you can run today

These are real pipelines running against live corpora. Not mockups, not wireframes — actual agent definitions you can create in the Agent Builder.

1 search GDPR data processor obligations
2 chat Extract key regulatory requirements
3 chat Synthesize compliance brief with citations
4 transform Format HTML digest
5 email Send to legal@company.no

GDPR Research Bot

Searches the Norwegian Legal Intelligence corpus in deep mode — three search passes that widen the query space, then cross-reference the passages. The AI extracts regulatory obligations, structures them into a compliance brief with exact source citations, and emails the digest every Monday morning.

In testing, the full 5-step deep pipeline completes in under 10 seconds against a corpus of 280,000+ legal document chunks.

  • Norwegian Legal Intelligence corpus (employment law, GDPR, AI regulation)
  • Deep mode: 3 search passes with query refinement
  • Source citations from EU directives and Norwegian statutes
1 search Force majeure, liability, IP clauses
2 chat Analyse changes and implications
3 transform Format compliance alert
4 email Send alert to legal team

Contract Clause Monitor

Monitors three high-risk clause categories — force majeure, liability caps, and IP assignment — across your contract corpus. Uses the iterate search mode to run each topic as a separate query, then analyses the combined findings for regulatory exposure.

Designed for weekly scheduled runs. When the corpus is updated with new contracts, the next scheduled run picks up the changes automatically.

  • Multi-topic iterate search (up to 5 topics per run)
  • Flags new or changed clause language
  • Automatic email alerts with findings summary
1 search Competitive product announcements
2 chat Identify top 3 competitive threats
3 transform Format executive summary
4 webhook POST to Slack #market-intel

Weekly Market Brief

Searches your market intelligence corpus for competitive product announcements, analyses trends and identifies the top three threats, then posts a concise executive summary directly into your Slack channel via incoming webhook.

The webhook step posts structured JSON that Slack renders as a rich message with sections and bullet points. No email, no manual distribution — it arrives where the team already works.

  • Direct Slack integration via webhook
  • Structured JSON payload for rich formatting
  • Schedule: every Monday at 08:00 Europe/Oslo

Agent Builder scales with your CaveauAI subscription

Agent limits grow with your plan tier. All tiers include scheduling, email delivery, webhook integration, and full run history.

Starter
3 agents

100 runs per month. All templates, scheduling, and email delivery included.

Start Trial
Enterprise
10 agents

500 runs per month. Full API access, custom pipelines, and dedicated support.

Start Trial
Beta
10 agents

500 runs per month. Full access for early testers and integration partners.

Request Beta

Build your first agent

Pick a template, point it at your corpus, and run it. See how it works before you commit to anything.

Open Agent Builder Read the User Guide
AI Chat — Beta Testing, Online Soon