JFYI
Just For Your Information — a passive MCP server and analytics platform that profiles developer coding habits and feeds profile-guided hints back to AI agents.
v2.12.0 MCP FastAPI SQLite Python 3.12+ Apache-2.0
What it does
JFYI runs as a background MCP server alongside your AI coding assistant. It observes every interaction — what you asked, what the agent produced, and whether you corrected it — and builds a curated developer profile over time. At the start of each new session, the agent reads this profile and can immediately act in alignment with your preferences, without needing to re-learn them from scratch.
Core mission: Profile the human. Serve the profile back to the agent at session start. Every feature is evaluated against that test.
The system tracks friction (corrections, latency, edit volume), infers patterns, and exposes a curated set of rules — the developer's "constitution" — that the agent can read before its first response. It also ships a web dashboard for reviewing notes, composing rules, exploring analytics, and managing the profile.
Quickstart
# Container (multi-arch: linux/amd64, linux/arm64)
docker run -p 8080:8080 -v jfyi-data:/data \
ghcr.io/hlan-net/jfyi-just-for-your-information:latest
# Helm chart (OCI, Helm 3.8+)
helm install my-jfyi \
oci://ghcr.io/hlan-net/charts/jfyi-mcp-server \
--namespace jfyi-system --create-namespace \
--set persistence.size=2GiThe server runs on port 8080 in SSE mode (MCP + web dashboard). For IDE integration (stdio mode):
jfyi serve --transport stdio --data-dir ./dataExplore
Architecture →
Mission, write-raw/curate/read-curated pattern, memory tiers, anti-patterns.
MCP Tools →
Full reference for all MCP tools, resources, and the REST API.
Features →
All shipped features across six phases, with rationale and implementation notes.
GitHub →
Source code, issues, and discussions.
Container image →
Multi-arch image on GHCR.
Helm chart →
Kubernetes deployment with hardened Pod Security Standard.
Release history
| Version | Phase | Theme | Status |
|---|---|---|---|
v2.3.0 |
Phase 1 — Foundation | Progressive disclosure, payload minification, read-only injection, OAuth 2.1 + RBAC | Shipped |
v2.4.0 |
Phase 2 — Memory Architecture | Compiled view memory, context compaction, three-tiered memory, background summarization | Shipped |
v2.5.0 |
Phase 3 — Advanced Retrieval | Vector embeddings (ChromaDB), instruction-tool retrieval | Shipped |
v2.6.0 |
Phase 4 — Security & Hardening | Inline DLP / PII redaction, developer behavior analytics, rule synthesis, agent provenance | Shipped |
v2.7–2.9 |
Operational | ChromaDB pod extraction, PVC model cache, JWT rotation, configurable TTLs, admin About page | Shipped |
v2.9.0 |
Profile Architecture | Notes vs Rules — two-tier developer profile (raw notes + curated rules) | Shipped |
v2.11.0 |
Evidence & Docs | Per-rule provenance in synthesis, source-notes preview, durable architecture doc | Shipped |
v2.12.0 |
Phase 6 — Vibe Coder Optimization | Tiered profiling, positive reinforcement, semantic rule inference, vibe telemetry, friction clustering, agent warming | Shipped |
v2.13.0 |
Dashboard Hardening | Agent Analytics page — per-agent alignment comparison, correction rates, friction scores | Shipped |
v3.0.0 |
Phase 5 — Protocol Expansion | ACP and A2A cross-framework agent interoperability (blocked on spec stability) | Shelved |