Loki
Self-improving multi-platform agent harness with an optional Jev companion for typed judgments and per-session model routing.
Overview
Loki, from WunderCorp, is a self-improving AI agent with a built-in learning loop: it creates skills from experience, improves them during use, persists knowledge, searches its own past conversations and builds a model of the user across sessions. It runs on a cheap VPS, GPU cluster or serverless backends, talks through Telegram, Discord, Slack, WhatsApp, Signal and the CLI from one gateway, schedules automations, spawns subagents, and works with any model via OpenRouter, OpenAI or custom endpoints.
Per the listings, Loki optionally adds a TypeSafe Jev companion that provides typed Choice, Score and Noul judgment tools to the agent and routes a new session to a model within the selected gateway. The README excerpt describes the general agent rather than the Jev integration, so its exact questions are not confirmed here. Documentation is available in several languages.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which model in the gateway should handle this new session?(inferred) | — |
| Other | Typed Choice, Score and Noul judgments exposed as agent tools | — |
How it works
- User starts a session on any platform
- Optional Jev companion routes the session to a model
- Agent may call Jev judgment tools during work
- Learning loop updates skills and memory
Strengths
- Optional, so the agent works without Jev
- Typed judgments available as tools
- Broad platform support
Caveats
- Jev integration details not in the README excerpt
- Large general-purpose agent
Getting started
Install @wundercorp/loki from npm and run loki model to choose a provider.At a glance
- Type
- Service
- Status
- beta
- Language
- TypeScript
- Runtime
- agent harness and gateway
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Loki optionally adds Jev typed-judgment tools and routes a new session to a model within the selected gateway.
logicrw/awesome-jev-projectsSelf-improving agent harness with an optional TypeSafe Jev companion for typed Choice, Score, and Noul judgments.
hellogumbo/awesome-jev
Frequently asked questions
How does Loki use Jev?
As an optional companion offering typed judgment tools and session model routing, per the listings.
Is Jev required?
No, it is optional.
Which platforms does Loki run on?
Telegram, Discord, Slack, WhatsApp, Signal and the CLI.
Related projects
One skill and MCP server for Claude Code, Codex and Hermes where Jev gates large-model turns, routes models and guards Git steps.
MCP serverbetaPythonSkill pack handing an agent's small decisions to Jev: model routing, skill choice, search, memory, compaction and computer use.
Agent skillbetaPythonPython harness for open-ended goals: AI-authored tools and checks, Jev choosing the next tool and arguments from fresh state.
ProjectbetaPythonRuns Claude Code on Opus 5.5 with Jev re-picking reasoning effort after every tool batch, via per-message effort so the cache survives.
ProxybetaTypeScriptFourteen opt-in Claude Code and Codex skills where Jev makes one bounded choice: which skill, model, test, file or next step.
ProjectbetaPythonTerminal coding agent where an LLM reads, edits and runs code, and Jev makes a few optional side decisions.
ProjectbetaTypeScript