How do I install the Zep CLI skill?

Run npx skillmds@latest add idanbeck/zep-cli in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Zep CLI safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: makes network calls, reads secrets. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Zep CLI?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Zep CLI free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Zep CLI?

idanbeck (@idanbeck) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Zep CLI by idanbeck — Agent Skill

Name: Zep CLI
Rating: 4.666666666666667 (3 reviews)
Author: idanbeck

Zep CLI

Use when operating the ZeP (Ze Experiments Platform) through its first-party `zep` CLI — authenticating (local or "Sign in with Zerg" SSO), minting/listing/revoking scoped API tokens, inspecting experiments/runs/datasets/backends, submitting runs, and reading the integration API discovery doc.

idanbeck Updated Sep 9, 2026

File contents

ZeP CLI

Use the zep CLI instead of scraping the ZeP web UI when you need to inspect or operate experiments, runs, datasets, scrapers, backends, storage, or API tokens.

Defaults

Run from /Users/idanbeck/zerg-ze/ze (invoke with poetry run zep ..., or zep ... if installed on PATH).
Config is ~/.zep/config.json; environment overrides are ZEP_API_URL, ZEP_TOKEN, ZEP_API_KEY, ZEP_WORKSPACE_ID, ZEP_CONFIG_PATH (env wins over the saved file).
Default API base is http://localhost:8080. Set once with zep config set-api-url <url>.
Prefer a scoped zep_ API token for automation; use zep auth login for interactive/operator tasks.
Many commands accept --json or already print JSON; prefer JSON for scripting.

Authenticate

Local credentials:

zep auth login you@example.com           # prompts for password
zep auth whoami

zep auth login you@zergai.com --sso       # prompts for your zergai.com password

Mint a scoped automation token

zep auth token create \
  --name agent \
  --workspace 1 \
  --scopes experiment:read,run:read,dataset:read,run:write \
  --expires-in-days 30
# The raw zep_... token is printed ONCE. Store it, e.g.:
export ZEP_TOKEN=zep_xxxxxxxx

List / revoke tokens:

zep auth token list
zep auth token revoke <id>

Inspect & operate

zep discovery                 # integration API capabilities (version, scopes, resources)
zep experiments list
zep runs list
zep runs submit <experiment> --backend modal
zep datasets list
zep backends list

Target a specific workspace with --workspace <id> or ZEP_WORKSPACE_ID. A workspace-pinned token ignores other workspace selections.

Scopes

Scopes are <resource>:<action> (e.g. run:read, experiment:write). A :write scope implies :read. Run zep discovery to see the authoritative list and each resource's required scopes. A token can never be minted with broader scopes than its creator holds.

Safety

Do NOT submit runs, cancel, or delete unless the user asks.
Show the target workspace before any mutating command.
Never echo a raw zep_... token into logs or chat; prefer export ZEP_TOKEN=... set out-of-band.
For agent automation, always use a narrowly-scoped zep_ token, never an operator session token.

---name: zep-clidescription: Use when operating the ZeP (Ze Experiments Platform) through its first-party `zep` CLI — authenticating (local or "Sign in with Zerg" SSO), minting/listing/revoking scoped API tokens, inspecting experiments/runs/datasets/backends, submitting runs, and reading the integration API discovery doc.--- # ZeP CLI Use the `zep` CLI instead of scraping the ZeP web UI when you need to inspect or operate experiments, runs, datasets, scrapers, backends, storage, or API tokens. ## Defaults - Run from `/Users/idanbeck/zerg-ze/ze` (invoke with `poetry run zep ...`, or `zep ...` if installed on PATH). - Config is `~/.zep/config.json`; environment overrides are `ZEP_API_URL`, `ZEP_TOKEN`, `ZEP_API_KEY`, `ZEP_WORKSPACE_ID`, `ZEP_CONFIG_PATH` (env wins over the saved file). - Default API base is `http://localhost:8080`. Set once with `zep config set-api-url <url>`. - Prefer a scoped `zep_` **API token** for automation; use `zep auth login` for interactive/operator tasks. - Many commands accept `--json` or already print JSON; prefer JSON for scripting. ## Authenticate Local credentials: ```bash zep auth login you@example.com # prompts for password zep auth whoami ``` Sign in with Zerg (SSO against zergai.com, if enabled on the server): ```bash zep auth login you@zergai.com --sso # prompts for your zergai.com password ``` ## Mint a scoped automation token ```bash zep auth token create \ --name agent \ --workspace 1 \ --scopes experiment:read,run:read,dataset:read,run:write \ --expires-in-days 30 # The raw zep_... token is printed ONCE. Store it, e.g.: export ZEP_TOKEN=zep_xxxxxxxx ``` List / revoke tokens: ```bash zep auth token list zep auth token revoke <id> ``` ## Inspect & operate ```bash zep discovery # integration API capabilities (version, scopes, resources) zep experiments list zep runs list zep runs submit <experiment> --backend modal zep datasets list zep backends list ``` Target a specific workspace with `--workspace <id>` or `ZEP_WORKSPACE_ID`. A workspace-pinned token ignores other workspace selections. ## Scopes Scopes are `<resource>:<action>` (e.g. `run:read`, `experiment:write`). A `:write` scope implies `:read`. Run `zep discovery` to see the authoritative list and each resource's required scopes. A token can never be minted with broader scopes than its creator holds. ## Safety - Do NOT submit runs, cancel, or delete unless the user asks. - Show the target workspace before any mutating command. - Never echo a raw `zep_...` token into logs or chat; prefer `export ZEP_TOKEN=...` set out-of-band. - For agent automation, always use a narrowly-scoped `zep_` token, never an operator session token.

Frequently asked questions