Stigmergy-mcp

The O(N) curve is a hypothesis: it holds while per-agent retrieval stays bounded. Phase 1 of the companion benchmark measured retrieved content per agent rising from about 0.6k to 3.7k tokens between 5 and 10 agents.
Power your agent
การติดตั้งจากผู้เผยแพร่ ไม่ใช่เอนด์พอยต์ของ PlugMCP
URL แบบรีโมตหรือคำสั่งบนเครื่องของคุณ? การติดตั้งสองแบบต่างกันอย่างไร
รันบน PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
เปิดจากผู้เผยแพร่

The O(N) curve is a hypothesis: it holds while per-agent retrieval stays bounded. Phase 1 of the companion benchmark measured retrieved content per agent rising from about 0.6k to 3.7k tokens between 5 and 10 agents.
MCP server implementing digital pheromone-based indirect coordination for AI coding agents
An MCP-compatible trace store for experimenting with indirect shared-state coordination among AI agents.
stigmergy-mcp is a coordination primitive rather than an agent framework, orchestrator, or runtime. Agents can deposit typed traces into a shared environment, retrieve traces near a task area, reinforce or weaken existing traces, and inspect high-intensity signals across a broader area.
Trace-based coordination is a hypothesis under evaluation, not a proven solution to multi-agent scaling.
Under cumulative, non-summarizing, sequential handoff with approximately constant predecessor payloads, repeated context transfer can grow quadratically with agent count. Under bounded trace retrieval, bounded retrieved volume, and bounded operation frequency, trace-based coordination can grow approximately linearly. Those forms are conditional. Message passing can summarize, cache, prune, and route selectively, while trace size, retrieval breadth, mechanism overhead, and fidelity loss can grow with system scale.
The companion Stigmergy-mcp-benchmark provides exploratory measurements. The current evidence does not establish a general crossover, matched-fidelity superiority, or a mechanism-specific benefit from decay and reinforcement.
Trace coordination may be worth evaluating when:
Direct handoff, bounded summarization, or a generic shared workspace may be preferable when:
Stigmergy describes indirect coordination through state left in a shared environment. The implementation supports three trace types:
Traces carry intensity, tags, metadata, and an exponential decay horizon. Other agents retrieve traces by area and effective intensity rather than by replaying a full transcript.
npm install stigmergy-mcp
git clone https://github.com/calabamatex/Stigmergy-mcp.git
cd Stigmergy-mcp
npm install
npm run build
# From npm
claude mcp add stigmergy -- node node_modules/stigmergy-mcp/dist/src/index.js
# From source
claude mcp add stigmergy -- node dist/src/index.js
npm run inspect
Traces persist to ./stigmergy.db by default.
STIGMERGY_DB_PATH=/path/to/traces.db
STIGMERGY_DB_PATH=:memory:
The database is created on first use.
deposit_traceWrite a trace into the shared environment.
sense_environmentRead traces near a specified area.
Results are sorted by effective intensity.
reinforce_traceStrengthen or weaken an existing trace.
get_gradientReturn the strongest traces across a broad area.
Effective intensity follows:
effective = intensity * exp(-elapsed_hours / decay_hours)
With the default decay_hours=24, a trace retains approximately 37% of its original intensity after 24 hours and approximately 14% after 48 hours. Runs lasting only minutes are unlikely to exercise the decay mechanism materially. Claims about the value of decay or reinforcement therefore require long-horizon tests or explicit ablations.
Traces below the pruning threshold are removed during deposit operations and are excluded from reads below the applicable query threshold.
Agent A:
deposit_trace(area: "src/auth/session.ts",
action: "found XSS in session handler",
trace_type: "danger",
intensity: 0.8,
tags: ["security"])
Agent B:
sense_environment(area: "src/auth/login.ts", radius: 1)
reinforce_trace(trace_id: "...", delta: 0.15)
Agent C:
get_gradient(area: "src/", limit: 5)
import { TraceStore } from 'stigmergy-mcp/store';
import { createServer } from 'stigmergy-mcp/server';
const store = new TraceStore('/path/to/traces.db');
const server = createServer(store);
See ARCHITECTURE.md for implementation details.
A writable shared trace store can become a prompt-injection, poisoning, and sensitive-data propagation surface. Production deployment should add controls appropriate to the environment, including:
The package should not be interpreted as a complete production security control plane.
The exploratory benchmarking harness lives at calabamatex/Stigmergy-mcp-benchmark.
npm run build
npm run dev
npm test
npm run test:coverage
npm run loc
npm run inspect
MIT
MCP server implementing digital pheromone-based indirect coordination for AI coding agents
เครื่องมือ
ยังไม่มีรายการเครื่องมือ เซิร์ฟเวอร์แบบรีโมตจะได้รายการหลังการตรวจสอบสถานะครั้งแรก
FAQ
คัดลอกคอนฟิก stdio สำหรับ Cursor, Claude Desktop, Claude Code หรือ VS Code ได้ในหน้านี้ คอนฟิกนี้ใช้การติดตั้งของผู้เผยแพร่ ไม่ใช่พร็อกซีของ PlugMCP
ระเบียนนี้ไม่ได้ระบุตัวแปรสภาพแวดล้อมที่จำเป็น
Stigmergy ทำงานแบบโลคัลผ่าน stdio
ตรวจสอบล่าสุดเมื่อ 2026-11-05 MCP server implementing digital pheromone-based indirect coordination for AI coding agents