zhoushoujianwork/easyeda-agent
zhoushoujianwork/easyeda-agent has 997 GitHub stars and 121 forks. GitHot measured +14 stars in 24 hours. Last checked 2026-11-09 23:10 UTC. Last push 2026-11-01.
嘉立创 EDA 专业版(EasyEDA Pro)开源 AI 自动化:原理图、PCB 布局布线与工程检查,支持 CLI / Agent Skill / MCP。配套 EdaDesk Desktop 内测版可下载体验。Open-source AI schematic and PCB automation for EasyEDA Pro.
Measured star growth
- in 24h
- +14
- in 7 days
- Update delayed
- in 15 days
- Update delayed
- in 1 month
- Update delayed
Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.
README.en.md
Original
easyeda-agent
Open-source AI automation for EasyEDA Pro: datasheets, schematics, PCB layout and verification.
Project website · Download Desktop beta · Releases · Plugin marketplace · 中文
easyeda-agent connects AI agents to EasyEDA Pro (嘉立创 EDA 专业版) through the official
eda.* API. It combines a Go CLI and local daemon, a connector extension, an Agent Skill,
and a local stdio MCP adapter. Actions use typed inputs and structured readbacks so agents
can inspect real components, pins, nets, geometry, DRC results and saved state.
Try EdaDesk Desktop
For an integrated workspace with projects, AI chat, schematics, PCB views and tool activity, download the companion EdaDesk Desktop beta →.
| Option | Best for | Start here |
|---|---|---|
| EdaDesk Desktop | A desktop workspace for projects, chat and tools | Downloads and installation |
| CLI / Skill / MCP | Existing Codex or Claude Code workflows, scripts and tool integrations | Install below · Quick Start · MCP setup |
Desktop 0.1.0 beta bundles easyeda-agent 1.16.0 and offers macOS Apple Silicon and Windows x64 packages through this project's public GitHub Release. No private-repository access is needed. See the Desktop website for current packages and supported platforms.
Desktop source remains private; an open-source date has not been decided. The engine in this repository remains available under its existing licenses. The macOS beta is ad-hoc signed and not notarized; the Windows beta is unsigned and has not been validated on a Windows machine. Connect your own installed and authenticated local agent or a custom model. Official Token grants and the unified model service are not open yet; downloading the app does not grant credits. Start with a disposable test project.
What it can do
| Workflow | Capabilities |
|---|---|
| Datasheet-driven libraries | Create Symbol, Footprint and Device assets from documented pin/package data; validate evidence and pin-to-pad mapping. AS07 case study |
| Schematics | Component selection, canonical pin/net data, module composition, placement, wiring, normalization and readback checks |
| PCB layout and routing | Module-aware placement, board outlines, keepouts, rule-aware short routes, vias, copper pours and silkscreen adjustment |
| Reusable circuits | Reuse circuit-block topology and inspect native CBB assets; follow the selection and reuse workflow |
| Verification and delivery | Connectivity and geometry checks, official DRC, BOM, netlist, manufacturing exports, design drawings, audit logs and explicit saves |
| Agent integration | CLI, Agent Skill and stdio MCP, including DeepSeek Harness setup |
See CLI Status, the feature inventory, and the schematic / PCB command references for current support and verification limits. Features do not imply that every generated board is production-ready.
Tell your agent what you need
After setup, ask a Skill-capable client to use easyeda-agent. For example:
Use easyeda-agent to inspect the current schematic and PCB. Read real components,
pins, nets, board outline, stackup and DRC rules first. Repair only issues supported
by that data. Read back the changes, check connectivity, geometry and DRC, save,
and list anything that still needs an engineering decision.
For a new device library, provide the exact part suffix and its datasheet. Ask the agent to search existing libraries first, verify pin/package evidence, validate the specification offline, then create and read back the assets. See the library-authoring workflow.
Examples and their acceptance boundaries:
- AS07 symbol, footprint and device.
- AT32F415 parameterized layout.
- ESP32-S3 four-layer board, historical case.
Install the open-source tools
Desktop users can follow the Desktop installation guide. For an existing AI client, install the CLI/daemon, connector and Skill:
1. Install CLI and Skill
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/zhoushoujianwork/easyeda-agent/main/install.sh | bash
Windows PowerShell 5.1 or PowerShell 7:
irm https://raw.githubusercontent.com/zhoushoujianwork/easyeda-agent/main/install.ps1 | iex
Both installers verify asset SHA-256, CLI version and Skill version before replacing
installed files. They detect supported AI clients and print the connector download URL.
Options include EASYEDA_VERSION, EASYEDA_INSTALL_DIR, EASYEDA_INSTALL_SKILLS,
EASYEDA_SKILL_PRESERVE and EASYEDA_GITHUB_PROXY.
On Windows the default CLI directory is %USERPROFILE%\.local\bin. The installer
prints PATH instructions; automatic user-PATH changes require -AddToPath when run
from a file, or EASYEDA_ADD_TO_PATH=1. It does not change machine PATH.
See Quick Start for manual installation, upgrades and troubleshooting.
2. Install the connector
Install EDA Agent Connector from the
official marketplace,
or import easyeda-agent-connector.eext from the
latest GitHub Release
using EasyEDA Pro's extension manager. Marketplace publication may lag behind GitHub;
check the actual installed components.
3. Open your project and check the connection
Enable Settings → Allow external interaction in EasyEDA Pro, then run:
easyeda health
The health report identifies daemon, connector, host, project and document readiness
and gives next steps. Commands start a missing daemon automatically unless
EASYEDA_NO_AUTOSTART=1 is set. Use easyeda update --check --exit-code to inspect
installed release alignment.
The project targets EasyEDA Pro V4, with V4.1.60 or newer V4 recommended. Host compatibility and action availability are separate checks. See V4 compatibility. Version diagnostics do not authorize design writes.
How it works
AI client / Agent Skill / stdio MCP
│
▼
easyeda CLI + local daemon
typed actions / audit / artifacts
│
▼
EDA Agent Connector (.eext)
official eda.* API
│
▼
EasyEDA Pro project
Inspect the exact target before mutations. Read back pins, nets and geometry after writes; explicitly save and reload at acceptance checkpoints. Screenshots and successful command exits alone do not prove a circuit is correct. See architecture and protocol.
Limitations
- Dense-board maze routing requires external Freerouting; the built-in router targets short, clear connections.
- The extension API has no programmable undo. Source checks, write guards, readbacks and failure rollback reduce risk.
- Dielectric thickness, Er and copper weight are not fully readable through the current API; automatic controlled-impedance approval is unsupported.
- Datasheet-based library creation needs exact part identity and adequate pin/package evidence. Missing evidence requires clarification.
- Software checks and historical examples do not establish physical, assembly or manufacturing acceptance.
Develop and contribute
- Public PCB solver packages and contribution plan: run
make pcb-solver-testoffline. - Circuit-block contribution guide.
- Documentation index, development environment and repository rules.
- Public Agent Skill.
Development targets dev; validated changes are promoted to main.
Thanks to EasyEDA Pro for its official extension API, @jlceda/pro-api-types, Freerouting, and the official EasyEDA extensions.
License
MIT. Files ported from
Easy_EDA_PCB_Beautify
under extension/src/beautify/ retain Apache-2.0; see NOTICE.
These engine licenses do not grant a license to the private Desktop source.
Star History
Explore related projects
affaan-m/ECC
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
300,686 starsNousResearch/hermes-agent
The agent that grows with you
262,395 starsmultica-ai/andrej-karpathy-skills
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
231,802 starsfirecrawl/firecrawl
Supercharge your AI agents with data from the web and beyond. Building the library for superintelligence. 🔥
200,444 stars