[ Guide ] # Jev in Claude Code: what you can build Jev in Claude Code: install the TypeSafe plugin, run /typesafe:typesafe-ai, add a permission hook and build agent checks. Checked 25 October 2026. Published 25 October 2026By Jev Hub [Back to Learn](https://jevhub.io/learn) [ TLDR ] - Jev cannot be the model that writes code in Claude Code, but it works beside it in three ways. - TypeSafe's plugin teaches Claude Code to write programs that call Jev, a permission hook lets Jev approve routine shell commands, and Jev can act as a fast check inside agents you build. - The plugin installs with two commands and runs as /typesafe:typesafe-ai. - Every figure from TypeSafe's and OpenRouter's cookbooks is the company's own. ## Jev in Claude Code, at a glance | | What to do | Where it is documented | | --- | --- | --- | | Install | claude plugin marketplace add typesafe-ai/skills, then claude plugin install typesafe@typesafe-ai | TypeSafe's agent-skill page | | Use it | Type /typesafe:typesafe-ai, or ask for the TypeSafe skill in your prompt | TypeSafe's agent-skill page | | Update | claude plugin marketplace update typesafe-ai, then claude plugin update typesafe@typesafe-ai; restart or run /reload-plugins | TypeSafe's agent-skill page | | Approve shell commands with Jev | A PermissionRequest hook in .claude/settings.json: matcher Bash, timeout 15, running a Bun script | OpenRouter's coding-agent cookbook | | Keys | TYPESAFE_API_KEY for the skill's experiments; OPENROUTER_API_KEY for the hook | Both pages | | Cost of the hook | About $0.0000168 per approval request, measured on a request of about 400 input tokens | OpenRouter's coding-agent cookbook | Commands and settings are from TypeSafe's agent-skill page and OpenRouter's cookbook, read on 25 October 2026. Check the linked pages for current commands. ## Can Jev work with Claude Code? Yes, beside it, not inside it. TypeSafe's documentation says Jev is not a coding-agent language model: it returns a structured answer (a choice with probabilities, a score on a rubric you define, or a yes or no probability) and writes no text or code, so it cannot replace the model behind Claude Code. What it can do is help in three ways: teach Claude Code to write programs that call Jev, act as a fast judge on the commands Claude Code wants to run, and sit inside agents you build as a cheap check between steps. The sections below take them in turn. The overview of every agent is on the main page. - [Jev in Codex: set it up and trust the hook](https://jevhub.io/jev-codex) - [Jev in OpenCode: skill, approvals and Zen](https://jevhub.io/jev-opencode) ## How do I install Jev in Claude Code? Claude Code installs TypeSafe's skill as a plugin. Run both commands, one after the other. TypeSafe's page warns against mixing install methods, because two methods leave duplicate copies. Claude Code · install, step 1 ``` claude plugin marketplace add typesafe-ai/skills ``` Claude Code · install, step 2 ``` claude plugin install typesafe@typesafe-ai ``` To use it, type the command below, or name the skill in your prompt, for example "use the TypeSafe skill". Skills run with Claude Code's permissions, so read the skill file before you install it. Claude Code · use the skill ``` /typesafe:typesafe-ai ``` To update, run these two commands, then restart Claude Code or run `/reload-plugins`: Claude Code · update, step 1 ``` claude plugin marketplace update typesafe-ai ``` Claude Code · update, step 2 ``` claude plugin update typesafe@typesafe-ai ``` ## What does the TypeSafe plugin do in Claude Code? It gives Claude Code the context to design Jev integrations. The [typesafe-ai/skills](https://github.com/typesafe-ai/skills) repository, MIT-licensed and showing 2.8 thousand stars when we read it, contains one skill, `typesafe-ai`. It covers how to describe the state, how to write questions, how to compose a request, how to probe live and where to find the current documentation. TypeSafe's page suggests three kinds of prompt. Ask Claude Code to find places where one judgment could replace complex parsing or fragile code. Export `TYPESAFE_API_KEY` and ask it to run cheap test queries and propose changes. Or point it at a TypeSafe cookbook and ask which pattern would make your code less fragile. - [How to write questions Jev answers well](https://jevhub.io/resources/question-design) ## Can Jev approve Claude Code's permission prompts? OpenRouter's cookbook shows how. A script runs on every `PermissionRequest` for the `Bash` tool, which you register in `.claude/settings.json` with the command `bun ./path/to/jev-permission-hook.ts` and a timeout of 15. A fixed risk list in code goes first: recursive deletes, force pushes, `sudo`, publishing, deploy commands and credential paths such as `.env` or `.ssh` always go to the normal prompt and Jev is never called. Everything else goes to Jev as two yes or no questions: is the command reversible and kept inside the project, and does it serve the task. The command is approved only if both scores are at least 0.9. Any error, timeout or unclear answer leaves the normal prompt in place. The cookbook measured about $0.0000168 for a request of roughly 400 input tokens, or about $0.017 per 1,000 checks. Three Claude Code details from the cookbook matter. The hook receives Claude Code's one-line description of the command in place of a task, so the second question checks the command against that description. The hook does not run where no prompt can appear: in `claude -p` and in background subagents, a request with no hook decision is denied. And deny and ask rules in your settings still beat an allow from the hook. ## What can I build with Claude Code and Jev? Three uses that TypeSafe and its partners have published, each with the numbers they report. All figures are theirs and not independently checked: - **Pick the right skill before each turn.** TypeSafe's skill-suggestion cookbook runs two Jev requests before every turn and adds the best skill to the prompt. On 488 test requests over 182 skills, with a Claude Haiku 4.5 agent, it reports wrong skill loads falling from 16.8% to 7.3% and needless loads from 9.8% to 4.0%. The run used Jev 1.12 and is dated 31 July 2026 in the cookbook. - **Decide whether an agent should take another step.** Vercel's published example asks Jev "should the agent take another step?" as a yes or no question and continues when the probability is at least 0.8. It is a cheap stop-or-go check between steps. - **Let an agent propose better questions.** TypeSafe's feature-discovery cookbook has an AI agent rewrite Jev questions over five rounds. On 2,000 wine reviews it reports an error of 1.772 points against 2.466 for word counts, using Claude Sonnet 5 as the proposer. - [Jev builds for coding agents](https://jevhub.io/builds/topic/coding-agents) ## What goes wrong in Claude Code? TypeSafe's page and the cookbook list these snags: - **The skill is not used.** Invoke it explicitly with the command above, confirm the installer targeted Claude Code, then restart. - **Claude Code invents request fields.** The plugin may be out of date. Update it and try again. - **Routing is off.** Move the thresholds: too high causes missed cases and too low causes false alarms. Make each question narrower. - **The hook approves something you would decline.** Add that command to the fixed risk list first. Jev sees only the request text, so the list, not Jev, is the real safety boundary. - **Scores move a little between identical requests.** Start at 0.9 and lower toward 0.8 only if safe commands keep prompting. - [Where Jev struggles](https://jevhub.io/where-jev-struggles) ## Common questions Not as the model that writes code, because Jev returns structured decisions, not text. You can install TypeSafe's plugin so Claude Code writes programs that call Jev, use Jev as a permission-prompt judge through a hook, or call Jev from agents you build. Run these two commands one after the other. Then type /typesafe:typesafe-ai or ask for the TypeSafe skill. Terminal ``` claude plugin marketplace add typesafe-ai/skills ``` Terminal ``` claude plugin install typesafe@typesafe-ai ``` Run these two commands, then restart Claude Code or run /reload-plugins. Terminal ``` claude plugin marketplace update typesafe-ai ``` Terminal ``` claude plugin update typesafe@typesafe-ai ``` OpenRouter's cookbook shows a PermissionRequest hook that sends safe-looking commands to Jev and approves them only when two yes or no scores are at least 0.9. A fixed list of dangerous commands is checked first and always goes to the normal prompt. No. OpenRouter's cookbook says the hook does not run where a prompt cannot appear, and in those cases a request with no hook decision is denied. OpenRouter's cookbook measured about $0.0000168 for one approval request of roughly 400 input tokens with Jev 1.13, or about $0.017 per 1,000 checks. Only commands that pass the risk list are billed. Installing the plugin costs nothing, but anything that calls Jev needs a key: a TypeSafe key for the skill's experiments or an OpenRouter key for the hook. ## Sources Written by Jev Hub from the sources below and checked against them on 25 October 2026. Where a figure is a company's own claim, the article says so. - [TypeSafe docs: the TypeSafe agent skill (read 25 October 2026)](https://docs.typesafe.ai/agent-skill) - [GitHub: typesafe-ai/skills (MIT licence, read 25 October 2026)](https://github.com/typesafe-ai/skills) - [OpenRouter cookbook: auto-approve coding-agent permission prompts with Jev (read 25 October 2026)](https://openrouter.ai/docs/cookbook/coding-agents/auto-approve-permission-prompts-with-jev) - [TypeSafe cookbook: skill suggestion (read 25 October 2026)](https://docs.typesafe.ai/cookbooks/skill_suggestion) - [TypeSafe cookbook: autoresearch feature discovery (read 25 October 2026)](https://docs.typesafe.ai/cookbooks/autoresearch_feature_discovery) - [Vercel changelog: AI Gateway supports TypeSafe clients and an HTTP API for Jev (21 September 2026)](https://vercel.com/changelog/ai-gateway-now-supports-typesafe-clients-and-an-http-api-for-jev) - [TypeSafe docs: Jev with coding agents (read 25 October 2026)](https://docs.typesafe.ai/introduction/coding-agents) ## Related content - [Jev in Codex](https://jevhub.io/jev-codex) - [Jev in OpenCode](https://jevhub.io/jev-opencode) - [Jev in Cursor](https://jevhub.io/jev-cursor)