OpenSpec
The leading spec-driven development workflow for AI agents. Implements the /opsx:propose → /opsx:apply → /opsx:archive cycle that scaffolds proposals, specs, design docs, and task checklists, then applies them and archives once shipped.
“OpenSpec is the spec-driven development workflow that AI coding agents have been waiting for. The pitch is straightforward: you tell the agent to propose a change (`/opsx:propose add-dark-mode`), and OpenSpec scaffolds a structured artifact bundle: proposal, specs, design, tasks. The agent then applies the tasks one at a time (`/opsx:apply`), and once done, archives the change with the specs updated for the next feature. The 43k stars are the loudest signal in this entire directory and they earn it. The architecture decision worth understanding: OpenSpec is built brownfield-first, scaling from a single developer to enterprise teams while preserving the team's existing workflow. Run it on a Greenfield repo and you get spec-driven development from day one; run it on an existing codebase and the proposals respect what already exists.”
INSTALL THIS SERVER
{
"mcpServers": {
"openspec": {
"command": "npx",
"args": [
"-y",
"@fission-ai/openspec@latest",
"mcp"
]
}
}
}
{
"mcpServers": {
"openspec": {
"command": "npx",
"args": [
"-y",
"@fission-ai/openspec@latest",
"mcp"
]
}
}
}
{
"mcpServers": {
"openspec": {
"command": "npx",
"args": [
"-y",
"@fission-ai/openspec@latest",
"mcp"
]
}
}
}
{
"mcpServers": {
"openspec": {
"command": "npx",
"args": [
"-y",
"@fission-ai/openspec@latest",
"mcp"
]
}
}
}
{
"mcpServers": {
"openspec": {
"command": "npx",
"args": [
"-y",
"@fission-ai/openspec@latest",
"mcp"
]
}
}
}
7 TOOLS AVAILABLE
OUR ASSESSMENT
- 43k GitHub stars and active maintenance signal real workflow adoption among teams using AI coding agents.
- Brownfield-first architecture; proposals respect existing code while shaping incremental change.
- Structured artifact output (proposal, specs, design, tasks) gives the agent a checklist to drive against, with clear hand-off points for human review.
- Zero-config quick start: `npm install -g @fission-ai/openspec@latest` then `openspec init`.
- Workflow conventions are opinionated; teams with established spec patterns may find friction adapting.
- New `/opsx:` workflow is the recommended path; older OpenSpec workflows exist but receive less ongoing development.
- Requires Node.js 20.19.0+; older runtimes fail at install.
OpenSpec runs locally and operates on the operator's filesystem. The agent reads and writes files under `openspec/` in the project directory plus generated artifacts under the project tree. Secrets and credentials stay outside OpenSpec's scope; spec-driven workflows execute the same code the agent was already authorized to write, so the existing project-level access controls apply. Production deployment of OpenSpec-generated changes uses the team's standard CI and code-review gates.
Teams adopting AI coding assistants who want repeatable, spec-driven feature workflows over prompt-and-pray. Works equally well on greenfield and existing codebases.
TECHNICAL DETAILS
ADOPTION METRICS
// Reading this43.6k stars place this as the most-starred entry in our directory. Star traction reflects strong adoption among teams running structured agent-driven development workflows.
// Reading thisTop-ranked dev-tools MCP by every stat we track. The brownfield-first architectural choice is the reason it scales beyond demo workflows.
SOURCES & VERIFICATION
We don't take any single directory's word for it. Before scoring, we cross-reference 5 public MCP sources, install the server ourselves against the clients we cover, and record when we last re-verified.
The same server, 5 different lenses. We reconcile these signals into our editorial score, which is why our number sometimes diverges from a directory-aggregate star count.
| Source | Their rating | Their star count | Their downloads | Last synced |
|---|---|---|---|---|
| AutomationSwitch This page | 4.5editorial | 43,650 | — | APR 28, 2026 |
| PulseMCP | — unrated | unavailable | unavailable | APR 28, 2026 |
| MCPMarket | — unrated | unavailable | unavailable | APR 28, 2026 |
| Official MCP Registry | — unrated | unavailable | unavailable | APR 28, 2026 |
| Awesome MCP Servers | — unrated | unavailable | unavailable | APR 28, 2026 |
| Smithery | — unrated | unavailable | unavailable | APR 28, 2026 |
// Counts are directory-reported; we don't adjust them. Discrepancies usually come from different snapshot times or star-caching.
OTHER DEV TOOLS MCP SERVERS
GitHub
Codebase awareness layer for accessing repositories, pull requests, issues, code search, and GitHub Actions. Fine-grained PAT scopes keep access minimal.
Chrome DevTools for Agents
Vendor-official Chrome DevTools MCP server from the Chrome DevTools team at Google. Lets agents control and inspect a live Chrome browser for automation, debugging, and performance analysis.
Terraform MCP Server
Vendor-official Terraform MCP server from HashiCorp. Integrates with the public Terraform Registry (providers, modules, policies) and HCP Terraform / Terraform Enterprise (workspace management, organisation listing, run management, private registry access).
Browserbase MCP Server
Vendor-built Browserbase MCP server combining cloud browser automation (Browserbase) with natural-language-to-action mapping (Stagehand). 6 tools: start, end, navigate, act, observe, extract. Available as a hosted endpoint at https://mcp.browserbase.com/mcp (LLM costs covered for Gemini) or self-hostable via @browserbasehq/mcp.
AWS Infrastructure as Code MCP Server
AWS Labs server for Infrastructure as Code lifecycle. Validates CloudFormation templates with cfn-lint, checks compliance via cfn-guard against AWS Guard Rules and Control Tower controls, troubleshoots failed deployments with pattern-matched analysis, searches CloudFormation and CDK documentation.
Context7
Documentation quality layer that fetches current, version-specific library documentation. Resolves library IDs and returns focused doc sections so your agent works with accurate API references.
DISCUSS YOUR
MCP REQUIREMENTS.
Evaluating a server, scoping an internal deployment, or working out whether MCP is the right fit at all. Start the conversation and we will point you at the right piece of the ecosystem.