#3232nd of 62 in Agent and MCP starters

echo-agent

Self-hosted Python agent runtime with layered memory and a Tauri desktop app

Stars
1k
License
MIT
Last commit
Nov 2026
Last release
Oct 2026
Language
Python
Stars, last 29 weeks+1,028
22 Apr 2026: 7 stars29 Apr 2026: 34 stars6 May 2026: 86 stars13 May 2026: 97 stars20 May 2026: 144 stars27 May 2026: 202 stars3 Jun 2026: 264 stars10 Jun 2026: 324 stars17 Jun 2026: 390 stars24 Jun 2026: 446 stars1 Jul 2026: 503 stars8 Jul 2026: 526 stars15 Jul 2026: 555 stars22 Jul 2026: 622 stars29 Jul 2026: 680 stars5 Aug 2026: 735 stars12 Aug 2026: 830 stars19 Aug 2026: 914 stars26 Aug 2026: 925 stars2 Sept 2026: 984 stars9 Sept 2026: 1,007 stars16 Sept 2026: 1,009 stars23 Sept 2026: 1,015 stars30 Sept 2026: 1,020 stars7 Oct 2026: 1,021 stars14 Oct 2026: 1,026 stars21 Oct 2026: 1,027 stars28 Oct 2026: 1,028 stars4 Nov 2026: 1,034 stars11 Nov 2026: 1,035 stars

7 on 22 Apr 20261,035 on 11 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Echo Agent is a Python 3.11+ agent runtime installed from PyPI, with a terminal mode, a Gateway serving a Dashboard and HTTP/WebSocket API, and message channels. It keeps Working, Episodic, Semantic and Archival memory with BM25 plus optional vector recall, and can generate skill candidates from execution traces. The repo also ships EchoAgent Desktop, a Tauri app with an embedded Rust engine, Theia IDE and browser and computer control.

Who it is for: Developers self-hosting agents with persistent memory or a local desktop agent

Strengths

  • Four-tier memory with BM25 and optional vector search fused by RRF
  • Skill evolution gates candidates through validation, risk checks and isolated evaluation, with manual rollback
  • Gateway offers Dashboard, HTTP/WebSocket API, CLI/TUI, webhooks and cron triggers
  • Tool calls pass permission policy and approval; supports MCP, Skills and Plugins

Weaknesses

  • README states current code is PolyForm Noncommercial; commercial use needs separate permission
  • Desktop and Runtime use different engines (Rust vs Python) with separate config, sessions and memory
  • Skill evolution is off by default and must be configured
  • Embedding the runtime is source-level only; README advises pinning versions

What it needs

  • Python, Remote model services via API key, Model routing with fallback
  • Needs Python 3.11+, LLM provider API key

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in Agent and MCP starters

See all 62
Also in Agent and MCP starters
RankProjectScore
1deepseek-harnessPlugin-based agent harness from DeepSeek with a local Web UI279.6k stars+32,096 (+13%) in 30 daysMIT98 out of 100
2composioTypeScript and Python SDKs that give agents authenticated tools across 1000+ apps31k stars+574 (+1.9%) in 30 daysMIT92 out of 100
3conductorDurable workflow engine that orchestrates microservices and LLM agents from JSON32.5k starsApache-2.091 out of 100
#4a2uiProtocol and renderers for agents to stream native UI instead of text Live demo ↗ (opens in a new tab)17.2k starsApache-2.087 out of 100
#5mcp-useTypeScript framework for building MCP servers with React Views and an Inspector Live demo ↗ (opens in a new tab)10.9k stars+132 (+1.2%) in 30 daysMIT87 out of 100