Agensi MCP

    1

    Use the Agensi skill marketplace from inside an agent: judge a skill, quote before spending, publish via MCP

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Agensi MCP

    Example session with this skill installed

    Find me a skill that writes cold emails, tell me what it costs, and only load it if I say yes

    • Read your context and instructions
    • Compiled the agensi mcp

    Searched Agensi for "cold email writing" — 4 candidates:

    1. Cold Outreach Sequences — best_for: multi-step B2B sequences — $4 one-time — 312 uses
    2. Cold Email Rewriter — best_for: tightening an existing draft — free — 89 uses
    3. Prospect Research Briefs — best_for: the research before the email — $2 — 145 uses
      Quoting candidate 1 before anything happens: nothing is charged at this step.
      It is $4.00, paid from your balance, and grants permanent ownership.
      Say yes and I call get_skill with confirm: true; otherwise I load the free one instead.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Agensi (https://www.agensi.io) is a marketplace of agent skills: a package with a SKILL.md an agent loads and follows, sometimes with scripts, sold once, by subscription, or free. This skill is the playbook for driving its remote MCP endpoint — https://mcp.agensi.io/mcp/auth, Streamable HTTP over OAuth 2.1, 25 tools — from inside a session.

    Once installed, the agent can do two things it could not before. Shop: search_skills returns ranked listings with the fields that separate them (best_for, requires, relevance, install_count, price); get_skill returns the actual SKILL.md to follow; my_library says what the account already owns so nothing is sold twice; my_credits reports the balance before a price is discussed. Spend safely: Agensi's own image, video, voice, music and data tools run on Agensi's keys through find_tools / skill_tools / quote_tool / run_tool / check_job, and the skill states the protocol as hard rules — quote first (nothing is charged), state the exact price and what it produces, wait for an explicit yes, then run with that quote id and that price. quote_status is checked before any quote is re-run, because a user can approve a quote in the tool card and that is how agents double-charge.

    The secondary workflow is publishing: creator_status as a readiness gate, submit_skill draft-then-confirm, update_skill with a version and changelog, upload_skill_media, set_skill_tools, edit_changelog, with listing state pollable through my_skills.

    It needs an Agensi account and an MCP client that can complete the OAuth authorization once — Agensi's own connector guide for Claude, ChatGPT, Cursor and Codex, or the DeepSeek Harness plugin, which runs the whole sign-in locally. Four reference files cover the tool surface, per-client setup, the creator flow and the failure signatures.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    It documents the Agensi MCP toolkit; it cannot install npm packages or DeepSeek Harness plugin bundles, and it will not buy a skill or run a paid tool call without an explicit yes from the user. A client that only accepts a URL plus a static header cannot complete the OAuth step on its own, so the initial authorization has to happen in a client that can, or through Agensi's connector guide. Review times, prices and category contents are Agensi's, so the skill reads them from the toolkit instead of stating them as fixed.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 6 days ago

    • Free with an Agensi account

    Accounts
    Agensi account
    Software
    MCP client with OAuth 2.1 supportNetwork access to mcp.agensi.ioInteractive sign-in oncePython
    Optional
    DeepSeek Harness with Web UIPaid balance for paid itemsCreator mode with payouts

    Permissions

    Allowed Hosts

    mcp.agensi.io
    www.agensi.io

    File Scopes

    ~/.dsh/dsh-runtimes/agensi-mcp/
    (token
    file
    is
    never
    printed)
    the
    skill
    package
    being
    published

    The Agensi token lives in a local file under $DSH_HOME and is never returned by any route, written to a log, or printed in chat; the skill forbids echoing it. Paid calls are quoted before they run and require an explicit yes.

    Listed6 days ago

    What's inside

    Frequently Asked Questions

    No. `quote_tool` prices a call and charges nothing, and `run_tool` requires the quote id plus the exact quoted price after the user says yes. Paid skills are the same shape: `get_skill` without `confirm` only reports the price.

    Call `quote_status` with the quote id. The user can press Approve on the quote card, which runs it without notifying the agent, and `quote_status` reports that as `already_ran` with the result.

    Nothing. The price is held on the balance and the actual cost is charged only on success, never above the quote; re-running the same quote id returns the same job rather than charging twice.

    No. Any MCP client that can reach `https://mcp.agensi.io/mcp/auth` and complete the OAuth 2.1 authorization works; the plugin is one install path that performs the sign-in locally.

    It is the alternative for when the package already exists on disk: files are handed over as text, a draft id comes back for review, and listing state is pollable with `my_skills`. Account and profile work still happens on the website.

    New in LLM Ops

    Free