Agentic Loop

Autonomous Agent Loop · Tool Use

The agent decides what to search, how many times, and when it has enough context. It loops freely until it signals completion with the finish tool.

Adds over Web-Augmented: Autonomous decision loop — the model chooses its own research strategy rather than running fixed queries. Each search informs the next. The agent stops when it decides it has enough context.
👤
User
Mercury Intelligence
Input
company name
⚙️
Flask API
POST /api/brief/v3
Orchestration
tools + prompt
Agent Loop
New
🧠
Claude Sonnet
Autonomous reasoning + tool calls
call
result
🔍
web_search
General search
📰
news_search
Recent news
💰
financials
Revenue · headcount
↺ loops until finish() called · max 6 rounds
finish(brief) + search log
👤
User
Mercury Intelligence
Receives brief
brief + sources + log
📋
Account Brief
Brief + sources + search log
Output
1
User submits a company name — model receives the task + all tool schemas
2
Model autonomously decides which tools to call and in what order
3
Results feed back into the model — it keeps searching until satisfied
4
Model calls finish() with the completed brief — search log returned for transparency