Daily MCP Registry Digest — 2026-03-26

by

Daily MCP Registry Digest (2026-03-26)

Overview

  • Total updated servers (multi-version observed in this data slice): 19
  • Deprecated servers: 0 (no entries marked deprecated)
  • Deleted servers: 0 (no deletion signals present in the provided data)

Note: This digest is based only on the supplied registry payload (100 records). “Updated” here means the same server name appears with multiple versions and a latest flag.

Notable changes (operator-relevant)

  • Transport / endpoint shifts
    • Adramp Google Ads (ai.adramp/google-ads) moved from sse to streamable-http between 1.0.0 and later releases; latest is 1.0.3 using https://mcp.adramp.ai.
    • OpenMandate (ai.openmandate/mcp) endpoint changed from https://api.openmandate.ai/mcp (0.3.1) to https://mcp.openmandate.ai/mcp (0.3.2+); latest is 0.5.1.
  • New/changed authentication expectations
    • Trends MCP (ai.trendsmcp/trends) adds a required Authorization: Bearer ... header in 1.0.1 (latest).
    • Rolli (ai.rolli/mcp) relaxes ROLLI_USER_ID from required (1.1.4) to optional with a default from 1.1.5 onward; latest is 1.1.7.
  • Rapid patch cadence / fast-moving clients
    • mcpcap (ai.mcpcap/mcpcap) advanced quickly from 0.6.4 through multiple releases to 0.9.6 (latest), indicating frequent changes that may affect tooling behavior and output stability.
    • Marketgenius (app.marketgenius/mcp) uses date-stamped versions with frequent publishes; latest is 2026-03-24.
  • Metadata / packaging correctness issues to notice
    • Kawa Code (ai.kawacode/mcp) has a mismatch in 5.0.3 where the npm package entry still lists version 5.0.2, then corrects to 5.0.4 in the latest. This can break automation that relies on package metadata for pinning.
    • SocialAPI (ai.social-api/socialapi) shows inconsistent server vs package versions across entries (server versions 0.1.0/1.0.0/1.0.1; npm package version listed as 1.0.1 throughout). Latest server version is 1.0.1, but operators should validate the package they install matches expected protocol behavior.
  • Minor functional/positioning updates
    • SimStim (ai.simstim/mcp) bumped to 1.0.1 (latest) with updated positioning (“Deploy sims to any screen”).
    • Himalayas (app.himalayas/mcp) evolved descriptions and added SSE alongside streamable HTTP; latest is 1.0.2.
    • Cookiy (ai.cookiy/cookiy) progressed to 1.7.9 (latest) and changed repository pointer earlier in the sequence, which may impact source review and supply-chain verification workflows.

Updated servers list (latest versions in this slice)

  • ai.adadvisor/mcp-server → 1.0.1
  • ai.adramp/google-ads → 1.0.3
  • ai.cookiy/cookiy → 1.7.9
  • ai.kawacode/mcp → 5.0.4
  • ai.marketcore/mcp → 1.0.2
  • ai.mcpcap/mcpcap → 0.9.6
  • ai.obris/mcp → 0.4.0
  • ai.openmandate/mcp → 0.5.1
  • ai.perplexity/mcp-server → 0.9.0
  • ai.rolli/mcp → 1.1.7
  • ai.simstim/mcp → 1.0.1
  • ai.social-api/socialapi → 1.0.1
  • ai.thinkneo/control-plane → 1.0.2
  • ai.trendsmcp/trends → 1.0.1
  • ai.preclick/preclick-mcp → 0.2.0 (note: parallel/related listing exists under ai.urlcheck/urlcheck-mcp)
  • ai.urlcheck/urlcheck-mcp → 0.1.5 (now branded as PreClick; “formerly URLCheck”)
  • app.cardog/mcp → 1.0.1
  • app.himalayas/mcp → 1.0.2
  • app.marketgenius/mcp → 2026-03-24

Why it matters for operators

  • Transport changes can be breaking: SSE → streamable HTTP shifts (e.g., Adramp) may require client configuration updates, proxy/LB tuning, and different timeout/stream handling semantics.
  • Endpoint moves require allowlist/DNS updates: host changes (e.g., OpenMandate api. → mcp.) can fail in locked-down egress environments until firewall and allowlists are updated.
  • Auth requirement changes impact uptime: newly required headers (e.g., Trends MCP) will hard-fail calls until secrets are provisioned and rotated through your secret manager and deployment pipeline.
  • Registry/package metadata mismatches break automation: discrepancies (Kawa Code, SocialAPI) can cause pinning errors, unexpected downgrades, or failed builds—validate installed artifact versions against the server entry you intend to run.
  • High release velocity increases regression risk: fast-moving components (mcpcap, date-stamped Marketgenius releases) should be staged with canarying, pinned versions, and explicit upgrade windows.

Leave a Comment