User repository on GitHub
Unofficial MCP server for Jev (Typesafe AI)
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Agents & MCP
Askjev — pZacca
An unofficial MCP server exposing one ask tool: your agent hands Jev some material and a list of plain questions, and gets typed answers with probabilities back in one round trip.
- Stars
- 1
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Author
- pZacca
- Last verified
- Oct 2026
This is one of several Jev listings for the same project. The canonical entry is Askjev — shantanugoel.
About this project
Unofficial MCP server for Jev, Typesafe AI's System One model. Not affiliated with Typesafe AI. Give your agent a fast second opinion. It asks a plain question about material it already has, Jev works out whether that is a yes/no, a scale, or a choice, and answers with calibrated probabilities. No generative model in the loop, one round trip, a fraction of the cost and latency of a sub-agent. An agent planning a workflow, asking Jev one question at a time. Every number is real.
Quoted from the project's README. Read the full README.
- Languages
- TypeScript
Highlights
- Jev decides for itself whether each question is a yes/no, a scale or a choice, so the agent never has to pick the primitive.
- Two Jev calls in the common case and three when a scale question arrives without options; more questions do not mean more calls.
- Every answer carries routing metadata, so a misrouted question type shows up instead of silently returning a wrong-shaped answer.
- Runs locally over stdio with npx or hosted on Cloudflare Workers, with configuration given for four named clients.
Watch out
This is an unofficial project and says so; it is not affiliated with Typesafe AI. The hosted endpoint builds a TypeSafe client from the key you send on every call and forwards your question, so if that matters to you use the local stdio path, which needs Node 22+ with TYPESAFE_API_KEY set. Claude Desktop is the one listed client that cannot use the hosted server, because its custom connectors cannot send an API key header. Output is deliberately raw: no threshold and no verdict, yes/no answers have no separate confidence because the probability is the signal, and the routing metadata is what tells you whether Jev picked the right question type. Scale questions arriving without options fall back to one of seven fixed five-level rubrics chosen in code, and that list is not configurable. The README is long and the digest is truncated, so this covers the tool, install, routing and rubric behaviour visible there; the error table is cut off. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
The hosted server runs this repository on Cloudflare Workers, and the README says it keeps nothing: every call builds a TypeSafe client from the key you send and forwards the question. That means your material and your key pass through a third-party endpoint on every call. The local path runs over stdio with the key in the environment and needs Node 22+. Claude Code and Codex were exercised end to end on both transports.
The README's claim is structural rather than measured: no generative model in the loop, one round trip, a fraction of the cost and latency of a sub-agent. What the routing documentation lets you count is calls, not money. Two in the common case, three when at least one question is a scale without options, and adding questions does not add calls because each step batches everything into a single request. Set the include options to usage if you want token counts back.
A second Jev call picks the closest built-in rubric from a fixed list of seven: intensity, quality, severity, likelihood, sentiment, frequency and agreement, each with five levels. The answer comes back as a fractional expected level with the full distribution, plus a note telling the agent to pass options for a rubric tailored to the question. The rubric list is hardcoded and not configurable, and option order matters when the options do form a scale.
No, and the README presents that as the point: there is no way to force the question type, so the agent does not have to think about it. With no options the router can return a yes/no, a score after picking a rubric, or an error asking for options, and a single-option question is rejected by schema validation before it reaches the router. The routing block reports how sure Jev was about the type and the rubric, so a wrong guess is visible rather than silent.
The failed question gets an entry with kind error and a message saying what to change, and the batch does not fail. Every other question is still answered, so one unanswerable prompt does not sink the call.
For a noul answer the probability of yes is the signal, so there is no separate confidence field; confidence appears on score and choice answers. Routing still exposes how sure Jev was about the question type.
ASKJEV_INCLUDE is a comma-separated list of routing (the default), usage for summed token counts, and model. Setting it to an empty string returns nothing but the answers, and the tool description says what the server emits.
More Jev resources like this
48GitHub stars
Agent harness for work that outlives one chat, with resumable turns, routed state machines, and observational memory across sessions and processes.
Jev Agents & MCP#community#jevbest#catalog-import
Community
dzhng/duet-agentTypeScript
dzhng/duet-agentTypeScript37GitHub stars
Assistant framework that stores specs, memory, and specialist agents as plain markdown files in your repo for Claude Code, Codex, Opencode, and similar runtimes.
Jev Agents & MCP#community#jevbest#catalog-import
36GitHub stars
Pi extension that strips the full Agent Skills catalog from every request and loads only the skills a Jev ranking call judges relevant.
Jev Agents & MCP#community#jevbest#catalog-import
102GitHub stars
Behavioral code search that asks Jev to score repository fragments for a plain-language question and returns exact excerpts with paths and lines.
Jev Agents & MCP#community#jevbest#catalog-import
Community
nassim-arifette/jevgrepTypeScript
nassim-arifette/jevgrepTypeScript18GitHub stars
Local engine with TypeScript and Python SDKs that runs Claude Code and Codex sessions as a warm team with a durable mailbox.
Jev Agents & MCP#community#jevbest#catalog-import
Community
masonlee39/Multi-AgentTypeScript
masonlee39/Multi-AgentTypeScript16GitHub stars
Single-command harness that turns a plain-language stack description into synced docs plus blocking hooks for Claude Code, Codex, and Pi, with deterministic presets or one Jev chooser call.
Jev Agents & MCP#community#jevbest#catalog-import
Community
kyu1204/oh-my-harnessTypeScript
kyu1204/oh-my-harnessTypeScriptFrom the Jev community
Posts from builders shipping with Jev right now.