LangGraph-powered shared brain
Build production-grade AI agents with traceable RAG
ContextRouter is a modular framework for multi-step agent workflows: retrieval, reranking, citations, tool orchestration, and security-aware provenance via Bisquit.
Prefer enterprise deployment & support? See ContextRouter Pro.
Quick feel
Python
from contextrouter.cortex import stream_agent
async for ev in stream_agent(
messages=[{"role":"user","content":"Explain RAG briefly"}],
session_id="demo",
platform="web",
):
if ev.get("event") == "text_delta":
print(ev["data"]["text"], end="")Streaming events, citations, and modular providers/connectors.
Designed for real systems
Clean separation of brain vs transport; strict contracts; pluggable modules.
Cortex orchestration
LangGraph state machine with nodes/steps split for maintainability and correctness.
RAG with provenance
Multi-source retrieval + reranking + citations, wrapped in Bisquit envelopes for traceability.
Provider freedom
Vertex AI, OpenAI, Anthropic, local models; Postgres/Storage; connectors like web/RSS/files.
Enterprise
ContextRouter Pro
Enterprise-first architecture with strong security boundaries, observability, and predictable scaling — designed to fit into existing platforms.
- Scalability: multi-instance safe state, shared backends, and operational guardrails for production workloads
- Traceability & observability: end-to-end provenance + structured traces for audits, debugging, and regression detection
- Advanced key management: KMS/HSM-backed signing (keys never leave KMS), rotation via `kid`, audited access
- Secrets & encryption posture: Secret Manager-based secrets, least-privilege access, break-glass controls and rotation playbooks
- Security hardening for RAG: guardrails against prompt injection + context isolation + safe logging and monitoring hooks
- Integration-ready: clean boundaries between brain and transport, plus plugin-based extensibility for existing systems
Discuss your project
Leave your email and a short message — we’ll follow up with next steps.
Best for: enterprise deployments, custom connectors/providers, security reviews.
Also: pilot projects and architecture consulting.