AgentScout Logo Agent Scout

MCP Ecosystem Weekly Snapshot — July 8, 2026

Weekly snapshot of the Model Context Protocol ecosystem: 550 repositories, key movers, new entries, and trend analysis for the week of July 2–8, 2026.

AgentScout ·
#mcp #model-context-protocol #dev-tools #weekly-tracker #ecosystem

Overview

The Model Context Protocol ecosystem continues its rapid expansion, reaching 550 tagged repositories this week — a gain of 39 new repos (+7.6% WoW). The top 30 projects alone account for 319,864 cumulative stars, with the official modelcontextprotocol/servers repository crossing the 88K star milestone.

Key metrics for the week of July 2–8, 2026:

MetricValueWoW Change
Total tagged repositories550+39 (+7.6%)
Top 30 cumulative stars319,864
Official servers repo stars88,169+561 (+0.6%)
New entries in top 3023
Python projects (top 30)10
TypeScript projects (top 30)9

Top 30 MCP Repositories

#RepositoryStarsWeek ChangeTrendLanguage
1modelcontextprotocol/servers88,169+561 (+0.6%)➡️ StableTypeScript
2DeusData/codebase-memory-mcp27,966+15,178 (+118.7%)🚀 SurgingC
3PrefectHQ/fastmcp26,022+260 (+1.0%)➡️ StablePython
4modelcontextprotocol/python-sdk23,554+129 (+0.6%)➡️ StablePython
5microsoft/mcp-for-beginners16,704+112 (+0.7%)➡️ StableJupyter Notebook
6modelcontextprotocol/typescript-sdk12,815+100 (+0.8%)➡️ StableTypeScript
7CoplayDev/unity-mcp12,201+1,232 (+11.2%)📈 RisingC#
8mcp-use/mcp-use10,264🆕 New🆕 New EntryTypeScript
9lastmile-ai/mcp-agent8,412🆕 New🆕 New EntryPython
10Upsonic/Upsonic7,902🆕 New🆕 New EntryPython
11grab/cursor-talk-to-figma-mcp6,882🆕 New🆕 New EntryJavaScript
12firecrawl/firecrawl-mcp-server6,872🆕 New🆕 New EntryJavaScript
13getsentry/XcodeBuildMCP6,042🆕 New🆕 New EntryTypeScript
14appcypher/awesome-mcp-servers5,672🆕 New🆕 New Entry
15MinishLab/semble5,538🆕 New🆕 New EntryPython
16holaboss-ai/holaOS5,490🆕 New🆕 New EntryTypeScript
17the-open-agent/openagent5,358🆕 New🆕 New EntryGo
18nanbingxyz/5ire5,265🆕 New🆕 New EntryTypeScript
19exa-labs/exa-mcp-server4,675🆕 New🆕 New EntryTypeScript
20WenyuChiou/awesome-agentic-ai-zh4,238🆕 New🆕 New EntryPython
21IBM/mcp-context-forge4,046🆕 New🆕 New EntryPython
22homeassistant-ai/ha-mcp3,858🆕 New🆕 New EntryPython
23IvanMurzak/Unity-MCP3,469🆕 New🆕 New EntryC#
24taylorwilsdon/google_workspace_mcp2,801🆕 New🆕 New EntryPython
25zcaceres/markdownify-mcp2,771🆕 New🆕 New EntryTypeScript
26bethington/ghidra-mcp2,720🆕 New🆕 New EntryJava
27stickerdaniel/linkedin-mcp-server2,681🆕 New🆕 New EntryPython
28genieincodebottle/generative-ai2,535🆕 New🆕 New EntryJupyter Notebook
29metatool-ai/metamcp2,502🆕 New🆕 New EntryTypeScript
30zinja-coder/jadx-ai-mcp2,440🆕 New🆕 New EntryJava

Notable Movers

DeusData/codebase-memory-mcp — +15,178 stars (+118.7% WoW)

The standout performer this week. This high-performance code intelligence MCP server, written in C, surged from ~12.8K to nearly 28K stars — more than doubling in a single week. The project indexes codebases into a persistent knowledge graph, enabling agents to navigate large codebases with minimal token consumption. Its C-based implementation offers significant performance advantages over Python/TypeScript alternatives for code search at scale.

mcp-use/mcp-use — New Entry at #8 (10,264 stars)

A fullstack MCP framework for building MCP applications compatible with ChatGPT, Claude, and other MCP servers. Its rapid ascent signals developer demand for unified frameworks that abstract away the complexity of building MCP-compatible tools and clients.

grab/cursor-talk-to-figma-mcp — New Entry at #11 (6,882 stars)

Grab’s official MCP integration between AI coding agents (Cursor, Claude Code, Codex) and Figma. This reflects the growing intersection of AI-assisted development and design tooling — enabling agents to read and manipulate Figma designs directly.

getsentry/XcodeBuildMCP — New Entry at #13 (6,042 stars)

Sentry’s official MCP server for Xcode build operations. A strong signal that major developer tooling companies are investing in MCP integrations for their core products, extending agent capabilities into iOS/macOS development workflows.

homeassistant-ai/ha-mcp — New Entry at #22 (3,858 stars)

Smart home meets AI agents. This unofficial Home Assistant MCP server enables agents to control and monitor smart home devices, pointing toward a future where MCP becomes the universal integration layer not just for developer tools but for IoT and consumer automation.

Ecosystem Breakdown

Language Distribution (Top 30)

LanguageCountShare
Python1033.3%
TypeScript930.0%
JavaScript26.7%
C#26.7%
Java26.7%
Jupyter Notebook26.7%
C13.3%
Go13.3%
Other/None13.3%

Python and TypeScript continue to dominate, collectively accounting for 63% of top projects. The presence of C (codebase-memory-mcp) and Go (openagent) in the top 30 signals that performance-critical MCP implementations are emerging beyond the scripting language mainstream.

Category Highlights

  • Enterprise: IBM/mcp-context-forge (4,046 stars) — AI Gateway for MCP/A2A protocol bridging; taylorwilsdon/google_workspace_mcp (2,801 stars) — full Google Workspace control via MCP
  • Gaming: Two Unity MCP servers (CoplayDev/unity-mcp at 12,201 + IvanMurzak/Unity-MCP at 3,469) — game development MCP tooling is maturing rapidly
  • Security/RE: bethington/ghidra-mcp (2,720 stars) + zinja-coder/jadx-ai-mcp (2,440 stars) — reverse engineering MCPs signal demand for AI-assisted security analysis
  • Web/Scraping: firecrawl/firecrawl-mcp-server (6,872 stars) + exa-labs/exa-mcp-server (4,675 stars) — web data access remains a top MCP use case
  • Smart Home: homeassistant-ai/ha-mcp (3,858 stars) — MCP expanding beyond developer tools into consumer IoT

🔺 Scout Intel: What Others Missed

The MCP ecosystem is undergoing a structural shift from “protocol infrastructure” to “application layer.” While most coverage focuses on the official SDKs and servers repo (which grew modestly at +0.6% WoW), the real story is the explosive growth of application-specific MCP servers. 23 of the 30 top projects are new entries this week — not because they were just created, but because the GitHub topic tagging and search index has matured enough to surface them.

Three underreported trends:

  1. Performance-native MCP is emerging. DeusData/codebase-memory-mcp’s 118.7% weekly surge (written in C) and the Go-based openagent suggest that the next wave of MCP innovation will come from performance-critical implementations, not just Python/TypeScript wrappers. This mirrors the early days of containerization, where the initial Docker ecosystem was Python/Node-heavy before Rust and Go implementations took over for production workloads.

  2. MCP is becoming the universal API bridge. The diversity of this week’s top 30 — spanning game engines (Unity), design tools (Figma), smart home (Home Assistant), security (Ghidra/JADX), enterprise SaaS (Google Workspace), and social platforms (LinkedIn) — indicates MCP is evolving from a “model context” protocol into a de facto universal tool-calling standard. The protocol’s simplicity (JSON-RPC over stdio/SSE) makes it trivially adoptable across domains that REST APIs or GraphQL would over-engineer.

  3. The aggregator layer is consolidating. Projects like metatool-ai/metamcp (MCP Aggregator/Gateway) and IBM/mcp-context-forge (MCP/A2A bridge) represent a new middleware tier. As the number of MCP servers grows past 500, the problem shifts from “how to build an MCP server” to “how to discover, orchestrate, and govern hundreds of them.” This is the infrastructure layer that will determine whether MCP scales to enterprise production or remains a developer novelty.

Looking Ahead

  • Watch for MCP specification updates as the protocol matures toward 1.0 — the current rapid ecosystem growth will likely trigger governance discussions
  • The aggregator/gateway tier (metamcp, mcp-context-forge) is poised for consolidation as enterprises need MCP management at scale
  • Security MCPs (ghidra-mcp, jadx-ai-mcp) may see accelerated adoption as AI-assisted security auditing becomes mainstream
  • The smart home / IoT segment (ha-mcp) could be the leading edge of MCP’s expansion beyond developer tooling into consumer applications

Data collected on July 8, 2026. Repository data sourced from GitHub API search (topic:mcp + topic:model-context-protocol, stars:>50). Week-over-week changes calculated against the June 24, 2026 snapshot.

MCP Ecosystem Weekly Snapshot — July 8, 2026

Weekly snapshot of the Model Context Protocol ecosystem: 550 repositories, key movers, new entries, and trend analysis for the week of July 2–8, 2026.

AgentScout ·
#mcp #model-context-protocol #dev-tools #weekly-tracker #ecosystem

Overview

The Model Context Protocol ecosystem continues its rapid expansion, reaching 550 tagged repositories this week — a gain of 39 new repos (+7.6% WoW). The top 30 projects alone account for 319,864 cumulative stars, with the official modelcontextprotocol/servers repository crossing the 88K star milestone.

Key metrics for the week of July 2–8, 2026:

MetricValueWoW Change
Total tagged repositories550+39 (+7.6%)
Top 30 cumulative stars319,864
Official servers repo stars88,169+561 (+0.6%)
New entries in top 3023
Python projects (top 30)10
TypeScript projects (top 30)9

Top 30 MCP Repositories

#RepositoryStarsWeek ChangeTrendLanguage
1modelcontextprotocol/servers88,169+561 (+0.6%)➡️ StableTypeScript
2DeusData/codebase-memory-mcp27,966+15,178 (+118.7%)🚀 SurgingC
3PrefectHQ/fastmcp26,022+260 (+1.0%)➡️ StablePython
4modelcontextprotocol/python-sdk23,554+129 (+0.6%)➡️ StablePython
5microsoft/mcp-for-beginners16,704+112 (+0.7%)➡️ StableJupyter Notebook
6modelcontextprotocol/typescript-sdk12,815+100 (+0.8%)➡️ StableTypeScript
7CoplayDev/unity-mcp12,201+1,232 (+11.2%)📈 RisingC#
8mcp-use/mcp-use10,264🆕 New🆕 New EntryTypeScript
9lastmile-ai/mcp-agent8,412🆕 New🆕 New EntryPython
10Upsonic/Upsonic7,902🆕 New🆕 New EntryPython
11grab/cursor-talk-to-figma-mcp6,882🆕 New🆕 New EntryJavaScript
12firecrawl/firecrawl-mcp-server6,872🆕 New🆕 New EntryJavaScript
13getsentry/XcodeBuildMCP6,042🆕 New🆕 New EntryTypeScript
14appcypher/awesome-mcp-servers5,672🆕 New🆕 New Entry
15MinishLab/semble5,538🆕 New🆕 New EntryPython
16holaboss-ai/holaOS5,490🆕 New🆕 New EntryTypeScript
17the-open-agent/openagent5,358🆕 New🆕 New EntryGo
18nanbingxyz/5ire5,265🆕 New🆕 New EntryTypeScript
19exa-labs/exa-mcp-server4,675🆕 New🆕 New EntryTypeScript
20WenyuChiou/awesome-agentic-ai-zh4,238🆕 New🆕 New EntryPython
21IBM/mcp-context-forge4,046🆕 New🆕 New EntryPython
22homeassistant-ai/ha-mcp3,858🆕 New🆕 New EntryPython
23IvanMurzak/Unity-MCP3,469🆕 New🆕 New EntryC#
24taylorwilsdon/google_workspace_mcp2,801🆕 New🆕 New EntryPython
25zcaceres/markdownify-mcp2,771🆕 New🆕 New EntryTypeScript
26bethington/ghidra-mcp2,720🆕 New🆕 New EntryJava
27stickerdaniel/linkedin-mcp-server2,681🆕 New🆕 New EntryPython
28genieincodebottle/generative-ai2,535🆕 New🆕 New EntryJupyter Notebook
29metatool-ai/metamcp2,502🆕 New🆕 New EntryTypeScript
30zinja-coder/jadx-ai-mcp2,440🆕 New🆕 New EntryJava

Notable Movers

DeusData/codebase-memory-mcp — +15,178 stars (+118.7% WoW)

The standout performer this week. This high-performance code intelligence MCP server, written in C, surged from ~12.8K to nearly 28K stars — more than doubling in a single week. The project indexes codebases into a persistent knowledge graph, enabling agents to navigate large codebases with minimal token consumption. Its C-based implementation offers significant performance advantages over Python/TypeScript alternatives for code search at scale.

mcp-use/mcp-use — New Entry at #8 (10,264 stars)

A fullstack MCP framework for building MCP applications compatible with ChatGPT, Claude, and other MCP servers. Its rapid ascent signals developer demand for unified frameworks that abstract away the complexity of building MCP-compatible tools and clients.

grab/cursor-talk-to-figma-mcp — New Entry at #11 (6,882 stars)

Grab’s official MCP integration between AI coding agents (Cursor, Claude Code, Codex) and Figma. This reflects the growing intersection of AI-assisted development and design tooling — enabling agents to read and manipulate Figma designs directly.

getsentry/XcodeBuildMCP — New Entry at #13 (6,042 stars)

Sentry’s official MCP server for Xcode build operations. A strong signal that major developer tooling companies are investing in MCP integrations for their core products, extending agent capabilities into iOS/macOS development workflows.

homeassistant-ai/ha-mcp — New Entry at #22 (3,858 stars)

Smart home meets AI agents. This unofficial Home Assistant MCP server enables agents to control and monitor smart home devices, pointing toward a future where MCP becomes the universal integration layer not just for developer tools but for IoT and consumer automation.

Ecosystem Breakdown

Language Distribution (Top 30)

LanguageCountShare
Python1033.3%
TypeScript930.0%
JavaScript26.7%
C#26.7%
Java26.7%
Jupyter Notebook26.7%
C13.3%
Go13.3%
Other/None13.3%

Python and TypeScript continue to dominate, collectively accounting for 63% of top projects. The presence of C (codebase-memory-mcp) and Go (openagent) in the top 30 signals that performance-critical MCP implementations are emerging beyond the scripting language mainstream.

Category Highlights

  • Enterprise: IBM/mcp-context-forge (4,046 stars) — AI Gateway for MCP/A2A protocol bridging; taylorwilsdon/google_workspace_mcp (2,801 stars) — full Google Workspace control via MCP
  • Gaming: Two Unity MCP servers (CoplayDev/unity-mcp at 12,201 + IvanMurzak/Unity-MCP at 3,469) — game development MCP tooling is maturing rapidly
  • Security/RE: bethington/ghidra-mcp (2,720 stars) + zinja-coder/jadx-ai-mcp (2,440 stars) — reverse engineering MCPs signal demand for AI-assisted security analysis
  • Web/Scraping: firecrawl/firecrawl-mcp-server (6,872 stars) + exa-labs/exa-mcp-server (4,675 stars) — web data access remains a top MCP use case
  • Smart Home: homeassistant-ai/ha-mcp (3,858 stars) — MCP expanding beyond developer tools into consumer IoT

🔺 Scout Intel: What Others Missed

The MCP ecosystem is undergoing a structural shift from “protocol infrastructure” to “application layer.” While most coverage focuses on the official SDKs and servers repo (which grew modestly at +0.6% WoW), the real story is the explosive growth of application-specific MCP servers. 23 of the 30 top projects are new entries this week — not because they were just created, but because the GitHub topic tagging and search index has matured enough to surface them.

Three underreported trends:

  1. Performance-native MCP is emerging. DeusData/codebase-memory-mcp’s 118.7% weekly surge (written in C) and the Go-based openagent suggest that the next wave of MCP innovation will come from performance-critical implementations, not just Python/TypeScript wrappers. This mirrors the early days of containerization, where the initial Docker ecosystem was Python/Node-heavy before Rust and Go implementations took over for production workloads.

  2. MCP is becoming the universal API bridge. The diversity of this week’s top 30 — spanning game engines (Unity), design tools (Figma), smart home (Home Assistant), security (Ghidra/JADX), enterprise SaaS (Google Workspace), and social platforms (LinkedIn) — indicates MCP is evolving from a “model context” protocol into a de facto universal tool-calling standard. The protocol’s simplicity (JSON-RPC over stdio/SSE) makes it trivially adoptable across domains that REST APIs or GraphQL would over-engineer.

  3. The aggregator layer is consolidating. Projects like metatool-ai/metamcp (MCP Aggregator/Gateway) and IBM/mcp-context-forge (MCP/A2A bridge) represent a new middleware tier. As the number of MCP servers grows past 500, the problem shifts from “how to build an MCP server” to “how to discover, orchestrate, and govern hundreds of them.” This is the infrastructure layer that will determine whether MCP scales to enterprise production or remains a developer novelty.

Looking Ahead

  • Watch for MCP specification updates as the protocol matures toward 1.0 — the current rapid ecosystem growth will likely trigger governance discussions
  • The aggregator/gateway tier (metamcp, mcp-context-forge) is poised for consolidation as enterprises need MCP management at scale
  • Security MCPs (ghidra-mcp, jadx-ai-mcp) may see accelerated adoption as AI-assisted security auditing becomes mainstream
  • The smart home / IoT segment (ha-mcp) could be the leading edge of MCP’s expansion beyond developer tooling into consumer applications

Data collected on July 8, 2026. Repository data sourced from GitHub API search (topic:mcp + topic:model-context-protocol, stars:>50). Week-over-week changes calculated against the June 24, 2026 snapshot.

3hu2efr5u318t8n6ui68ul░░░79cm89ybexqua4necaxx66du8z6c1dca████fho62wc4096qeaggwm00hhaqhs9p72fo7░░░09moaevz3krwvii632l4u1ug0lx9gsvch████wgv3ja5chbt3tj31z5m87iahw8ryn87n████sniaw0yvuk5sb4bsp1atozrc7yo5cgt░░░1d3b3lwbfr8ez8tuso327cy8tzkvreao████ajjc9vx1p7ise4eimio7hk49kotncl87████d1f7mv02w9vvae9w053cafm5zbipslhe████7fkj3sjt6jcga5fs3nuz4i54wmcz0qdsf░░░s3sm9wef6rcp2j4vbftnpr6btxek0m████d90fob9qwopsllz3g1d0tbbqa3ewojyre░░░7i62ynwxb5a1m6cx30jhcthuksbo9elmn████big4cbf6p991ksisci49lyjjcamcxq178a████6w0cz5bk53tpqd8zukaroogqgf7reydes████0dgf0uybqn55waf88rr33enb5kiztgui░░░g2jngp8bj68afz8kv66df6bd3fo22u99h░░░nosg4lfh4yclkm49fsdrj734agd0ecbd░░░3d0txpj1ig9dta7zms2epj6n54p2n46q░░░t0kmxeqboyfzaj1mifbr9grnt1zj7wk9f░░░z9qlokq62351wzv1qtsk6q97lipe7twj░░░ogv8hlcf91ponfxoltosykmgzzgl2q9he████r8v8ooklcivqahfwot2pq30h6jargm████y5zxx8tjdac860920ams4atm3t9s1044████76yb157u1vhrww1wantqffgl0ak29aqu████n7h5wevdhchdjzvros5exdb91gfunmg████nu0tfte0zxfy88jbehcjbikrjcurmplv████hl6vrt18bbj4879sm406kqzfm0im5035r░░░xsasyq0xz3msaj6w72nhynxxz4r439l1░░░fqrs92uporjgbst1tz9y2omz927gznn░░░8vmrvwo83gylohwfrvjeljxy22nyt4co████bm7rcn60m2x4zmhje3aeqwsm5019f14h░░░i4zfq1d1j6sd6c8nr3y6qsovu4csxmdn████wzgqb1iyaiodf0hd2nqyhcpxnqnqroa8░░░fv52k9dp5gg3gusc72ihmkjww2vo2ydml████qdpa4v5588d6zhshbzw3vcmrhw2kii9s████14yff04wnp78czoc8cctkphzj3dfuqcf░░░to04i80qicjx9hc2zfgyt8s992y3ezr7░░░05xf9rsx8kzsku5q2auntpg7cje0ce857l░░░unyrvj5sn8oopywsdqd3vhj3yo7d0937h████56ghyshaaq6gfs1e836oh75nzjo242skm████21ig7fpnhly70krd85pmvlcnk1s3qu16h████1qaqv1jnrk9jgu7phe1039fu5o6oy04wh████ygr0ebq308i8raqb1iphb5b6h8mhhomf░░░2mpw392picuqmutj2ktnms81n2a7hj8░░░vafltskjsuevraugrhfesdtwnn5wpag48░░░bi3p9739r94n3jrnsoiceocvz4jj5yyg████yz3apx6yzoa6km35pinlt7eegid5na4v░░░kyz3awfk7hpp1sqm8fp9yjemn09ji4jea░░░ukll3v952rpcqhjsu2vq6tn17fiaizjk░░░a2alyhe4nzs