Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+22 more

    swarm

    1

    A hierarchical multi-agent orchestration manual for cost-aware and privacy-safe task routing across LLM fleets.

    $9

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    swarm

    swarm

    Example session with this skill installed

    Create a swarm plan for a project that involves summarizing 50 sensitive internal documents (low complexity) and designing a secure API architecture (high complexity) with a $20 budget.

    • Read your context and instructions
    • Compiled the swarm

    I have generated a SWARM-X deployment plan based on your requirements. The plan optimizes for privacy by pinning the summarization tasks to your local fleet while reserving budget for the high-complexity architecture design.

    SWARM-X Deployment Plan: Project "Secure Architect"

    MetricValue
    Plan Score92/100
    Primary BottleneckNone (Budget healthy)
    Privacy Compliance100% (Sensitive tasks routed to Local)
    Est. Total Cost$4.20
    Savings vs Frontier-only78%

    Topology Assignment

    • L0 Strategic Coordinator: claude-3-5-sonnet (Cloud)
    • L1 Domain Directors:
      • Director-Research (summarization domain)
      • Director-Engineering (design domain)
    • L2 Specialist Agents:
      • 50x ollama-llama3-8b (Local) -> summarize_pii
      • 1x claude-3-5-opus (Cloud) -> design_api_spec

    Memory & Adaptive Config

    • CMG Scopes: session, task, agent (Local store enabled)
    • AMR Triggers: UPGRADE enabled if design_api_spec latency exceeds 5s; CHECKPOINT every 5 documents.
    • QA Gate: L2 self-check followed by L1 director spot-audit for the API design.

    Next steps

    1. Run python3 scripts/swarm_planner_report.py --data spec.json to view the full dark-mode visualization.
    2. Initialize the local CMG store using scripts/cmg_memory.py.
    3. Approve the L0 synthesis prompt to begin the intake stage.

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

    About this skill

    Monolithic agent swarms often suffer from static decomposition, ephemeral memory, and ballooning cloud costs. SWARM-X solves this by providing a hierarchical-mesh orchestration framework that routes subtasks across a heterogeneous fleet of local and cloud models based on cost, privacy, and complexity.

    What it does

    • Heterogeneous Routing dynamically assigns subtasks to local or cloud models to minimize cost while meeting privacy constraints.
    • Hierarchical Topology structures agents into L0 coordinators, L1 directors, and L2 specialists for governed fanning out.
    • Persistent Cognitive Memory manages cross-run memory scopes including session, task, agent reputation, and organizational standards.
    • Adaptive Recomposition triggers mid-run actions like agent spawning, upgrades, or merges based on emergent task complexity.
    • Deterministic Planning scores swarm configurations from 0-100 and flags the single largest bottleneck before you spend budget.

    How it works

    1. Define the Spec by listing subtasks with their complexity and privacy needs alongside your model fleet's cost and capabilities.
    2. Generate the Plan using the swarm_planner.py tool to compute the L0/L1/L2 topology and initial model assignments.
    3. Configure Memory and Triggers by setting CMG scopes and AMR health-check intervals for mid-run adaptation.
    4. Execute and Review the dark-mode HTML deployment report to verify budget fit and privacy compliance before sign-off.

    Frameworks & tools

    The toolchain is built for Python 3 using only the standard library. It includes dedicated scripts for planning, routing (DMR), memory management (CMG), and report generation, alongside an MCP server for agentic integration.

    Why this beats prompting it yourself

    Manual orchestration leads to "frontier-model tax" where expensive models perform trivial tasks. This skill automates the argmin cost-optimization and enforces hard privacy filters that manual prompts frequently ignore or leak.

    Use cases

    • Orchestrating a local-first research swarm that only uses cloud models for final high-reasoning synthesis.
    • Managing cost-constrained development agents where sensitive PII must stay on local Ollama instances.
    • Building long-running swarms that require persistent agent reputation and procedural memory across sessions.

    Known limitations

    The tools are deterministic and offline. They do not have live access to model pricing, current API latency, or real-time model quality scores.

    How to use

    Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scanned

    Verified clean 7 days ago

    • Passed all security checks, Safe to install

    Listed7 days ago

    What's inside

    Frequently Asked Questions