Next-Gen Orchestrationv2.0

Research that shows its work.
Plans you can execute.

Topic classification, source tiering, multi-source verification, and live-streaming traces working in harmony. A multi-agent framework built for developers who demand grounded, structured output.

Free Preview • No Credit Card Required • BYOK Optional • Credentials Locally Encrypted

The Problem We Fixed

We audited our early builds against research topics like Muhammad Ali Jinnah. Generic LLMs defaulted to unstructured boilerplate and treated forum postings identically to encyclopedic records. Here is the structure we rebuilt.

Topic Classification: biography (98%)Grounded & Verified

Muhammad Ali Jinnah: Political Leadership and the Founding of Pakistan

1. Early Life & BackgroundEducated in Karachi and London. Joined Lincoln's Inn as the youngest student called to the English bar [1].
2. Constitutionalist ReformerAdvocated Hindu-Muslim unity via the Lucknow Pact (1916). Resigned from the Congress Party due to shifts in protest policy [2, 3].
Verification Details:

Structured dynamically using the biography template. Excluded UGC forum entries. Multi-source claims verified against primary library indices before drafting the final text.

Verified Primary Sources:
Britannica Online [1]National Archives Office [2]Lincoln's Inn Records [3]

The 4 Specialized Agents

Four purpose-built agents operating within defined bounds to fetch, map, plan, and compile verified research.

Streaming

Research Agent

Input

Any research topic or comparative query

Process

Classifies into 1 of 8 templates, queries Serper, tiers & deduplicates sources, cross-verifies major claims

Output

Structured markdown report with inline citations, streamed live via SSE with PDF export

Launch Agent
Structured Output

Task Planner

Input

High-level project goals or roadmaps

Process

Leverages Qwen 3.6 27B via Groq with Pydantic JSON schema enforcement to build deterministic subtask hierarchies

Output

Hierarchical phase decomposition with interactive subtask states and agent assignments

Launch Agent
RAG Pipeline

Document Analyst

Input

Single or multi-document PDF collections

Process

Uploads to Cloudflare R2, parses using PyMuPDF, index chunk vectors into a composite FAISS store via Gemini embeddings

Output

Grounded citation-backed chat interface with direct snippet-level highlights

Launch Agent
Orchestrator

Supervisor

Input

Complex multi-layered developer objectives

Process

Coordinates specialized sub-agents using a LangGraph StateGraph state machine to trace workflow execution

Output

Live token synthesis combining research, structured plan, and document grounding in one trace

Launch Agent

Live Execution Trace

Verifiable reasoning, not a black-box spinner. Watch the Supervisor map, schedule, and orchestrate research traces in real time via Server-Sent Events.

supervisor_event_stream.log
Classifying
2
Searching
3
Reading & Tiering
4
Verifying
5
Writing
SSE Stream Connection Status: CONNECTED RECEIVING

Bring Your Own Key (BYOK)

We utilize a direct two-layer credential model. You get an authenticated free allocation every single day. But when you require industrial scale, simply provide your own Gemini or Groq API credentials. The caps dissolve immediately, and you pay providers directly.

Encryption Standard:

Encrypted at the application layer with per-user key derivation before it's ever stored. Never logged, never sent to the frontend.

DAILY CAPACITIES

Free PreviewUnauthenticated IP limit
5 Runs Total / day
Signed InAuthenticated user limits
Research & Planner: 15 / day
Docs & Supervisor: 10 / day
BYOK ConfiguredYour own provider credentials
UNLIMITED

Infrastructure Transparency

No hidden layers. Open architecture mapped directly to production-grade services.

VPS Backend Hosting

Central FastAPI server handling LangGraph orchestration pipelines.

Vercel Frontend

Static Next.js compilation optimizing client-side event listeners.

Supabase Data Layer

User profiles and authenticated sessions stored securely.

Cloudflare R2 Object Store

Documents and FAISS vector structures saved on remote S3 storage.

Upstash Redis Throttling

IP-based rate limits (10 requests/min) keeping nodes active.

8 Verified Test Suites

Automated test matrices covering auth, encryption, and export mechanisms.

Privacy Shield: Private research items return a generic 404 error rather than revealing ownership.Shared report linkages are explicitly opt-in.

Frequently Asked Questions

Everything you need to know about quotas, key privacy, and architecture.

Is it free?

Yes. Standard registration grants 15 research and planning runs daily. If you require unlimited capability, you can bring your own Gemini/Groq keys, which bypasses all caps.

Are my API keys safe?

Absolutely. Keys are encrypted at the application layer with a per-user key derivation salt (AES-128-CBC + PBKDF2). They never touch server logs, are never sent back to the frontend, and are used solely to authenticate your queries directly with providers.

How is this different from ChatGPT or Perplexity?

Unlike static LLM searches, AgenticEra dynamically classifies topics to choose optimized outline structures, filters out unverified user-generated content (UGC), cross-verifies claims across multiple sources, and links directly to structured execution planners.

Why is there no paid premium tier?

We prioritized engineering quality and performance over billing infrastructure. By offering Bring Your Own Key (BYOK) support, builders can run queries at actual API cost without middleman subscriptions.

LogoAgenticEra Studio

Built by Syed Muhammad Hassan • AgenticEra Systems • Powered by Gemini & LangChain