Back to Home

Meet the Agent Fleet

Six specialized AI agents work together to assist your research. Each has a defined role, logs its actions, and defers to you for judgment calls.

Architecture

Human-Governed AI Assistance

You Send a Message

Ask a question, request an analysis, or give a task to the Agent Fleet.

Agents Route & Execute

The right agent handles your request. Complex tasks may involve multiple agents.

You Review & Approve

Agent outputs require your review. Sensitive actions need explicit approval.

Agent Profiles

Six Specialized Agents

Each agent has deep expertise in a specific domain of research work.

Planner

Mission decomposition and task orchestration

The Planner agent breaks down high-level research goals into actionable tasks. It understands the stage machine and routes work to appropriate agents.

Example Prompts

  • "Break down a systematic review into sub-tasks"
  • "Create a timeline for meta-analysis completion"
  • "Prioritize literature search tasks"

Outputs

Task listsMission plansWorkflow diagrams

Literature

Paper search, extraction, and claim analysis

The Literature agent searches databases like PubMed, helps process uploaded PDFs, and extracts structured claims from papers.

Example Prompts

  • "Search for PD-1 inhibitor trials in NSCLC"
  • "Extract efficacy data from this paper"
  • "Summarize the methods section"

Outputs

Paper summariesClaim cardsCitation metadata

Analysis

Statistical methods and computational analysis

The Analysis agent helps design statistical approaches, run calculations, and interpret results. It logs all analytical decisions.

Example Prompts

  • "Calculate pooled effect size using random effects model"
  • "Run heterogeneity analysis (I², Q-test)"
  • "Suggest appropriate statistical tests"

Outputs

Analysis scriptsEffect estimatesStatistical reports

Manuscript

Document drafting and citation management

The Manuscript agent generates draft text, manages citations, and helps structure your final document according to journal requirements.

Example Prompts

  • "Draft an introduction for this meta-analysis"
  • "Format citations in Vancouver style"
  • "Generate a methods section from the protocol"

Outputs

Section draftsFormatted citationsDocument outlines

Reviewer

Critical evaluation and quality feedback

The Reviewer agent provides critical feedback on your work, identifies potential issues, and suggests improvements.

Example Prompts

  • "Review this analysis for methodological issues"
  • "Check if conclusions match the data"
  • "Identify potential biases in study selection"

Outputs

Review commentsIssue flagsImprovement suggestions

QC

Data validation and quality control

The QC agent validates data quality, checks for inconsistencies, and ensures outputs meet quality thresholds before proceeding.

Example Prompts

  • "Validate this dataset for missing values"
  • "Check consistency across extracted data"
  • "Verify calculations match source data"

Outputs

Validation reportsQuality flagsConsistency checks
Interaction

Real-Time Streaming

Agent responses stream in real-time through the Agent Fleet pane. See exactly what's happening as agents work on your requests.

Server-Sent Events (SSE)

Responses stream token-by-token as agents generate them. No waiting for complete responses—see progress immediately.

See the agents in action

Book a demo and interact with the Agent Fleet live.

Book a Demo