Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

rlcd-gateway

Self-hosted Go gateway for LLMs and decision models that prunes context, and audits and calibrates Jev or open-rlcd decisions.

Servicebeta
View on GitHub ↗

Overview

RLCD Gateway is a single Go binary that sits between apps and model providers. Claude Code, Codex, OpenCode and any OpenAI or Anthropic SDK app route through it to the provider you choose (Anthropic, OpenRouter, OpenAI, Groq, Ollama, vLLM and others) with aliases, per-request rules and fallbacks, while subscription logins are forwarded untouched. It prunes stale tool output in a prompt-cache-aware way (starting in shadow mode, with an MCP rlcd_recall tool to get content back), classifies and recovers provider errors, issues gateway keys with limits, and shows every call, cost and a context X-ray in a live dashboard.

For decision models it is a drop-in Jev-compatible proxy for POST /v1/systemone, fronting TypeSafe's Jev or self-hosted open-rlcd. Apps change only their base URL; answers pass through byte for byte, every call is logged, outcomes can be recorded to measure accuracy and calibration (ECE), a mirror checks parity between backends, and routing rules can depend on a decision's answer.

What the decision model decides

TypeQuestionAnswers
Yes / noExplicitly asks for a refund?—
OtherAny System One decision proxied for audit, calibration and routing—

How it works

  1. Run the gateway and point apps' base URL at it
  2. Proxy LLM calls with pruning and fallbacks, or System One calls to Jev/open-rlcd
  3. Log each decision and record outcomes
  4. Measure accuracy and ECE, compare backends, route on answers

Strengths

  • One base URL for many providers
  • Calibration tracking for decision models
  • Context pruning to cut token costs
  • No system proxy or certificates needed

Caveats

  • Pruning should be validated in shadow mode first
  • Calibration needs outcome data you record

Getting started

npx rlcd-gateway (or uvx rlcd-gateway), then ANTHROPIC_BASE_URL=http://127.0.0.1:4777 claude, or set the System One base URL to http://127.0.0.1:4777.

At a glance

Type
Service
Status
beta
Language
Go
Runtime
Local gateway + web dashboard
Has code
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does the gateway do for Jev users?

It proxies /v1/systemone calls, logs them, records outcomes for accuracy and calibration, and can route on answers.

Does it change Jev's answers?

No; the README says answers pass through byte for byte.

Does it work with Claude Code subscriptions?

Yes, logins are forwarded untouched.

Related projects

  • jev-gate-mcp

    MCP gateway that trims noisy tool output and offers task classification and review advice, splitting work between local Laya and Jev.

    MCP serverprototypeTypeScript
  • stuntd

    Local Jev-compatible proxy that records an app's typed decisions, distils a head per decision site on Laya, and answers when confident.

    ProxybetaPython
  • jev-pipeline-skill

    Agent skill and CLI that chains Jev question design: find decision calls, check fit, draft, lint, calibrate thresholds and plan a cascade.

    Agent skillprototypePython
  • jev-x-kit

    MCP server, CLI and Claude Code skill with Choice, Score and Noul primitives and a confidence gatekeeper, hosted or offline.

    MCP serverbetaTypeScript
  • JevT++

    C++20 library for typed routing, classification and scoring with compile-time enum schemas, local Laya and optional System One.

    LibrarybetaC++
  • jevcompat

    48-requirement spec and conformance suite for Jev-compatible servers, with a normalising proxy; shows confidence is not portable.

    CLIbetaPython