Agent Skills vs Model Context Protocol (MCP)
Agent Skills is a protocol from Anthropic (Apache-2.0 licence, remote hosting, 27k GitHub stars, last push 2026-11-04); Model Context Protocol (MCP) is a protocol from Agentic AI Foundation (Linux Foundation) (MIT AND Apache-2.0 licence, 9.7k GitHub stars, last push 2026-11-21). They share 0 of 1 capabilities and integrations identified for either (overlap 0.60); only Model Context Protocol (MCP) covers Authorization. Figures as of 2026-11-21. DER
Agent Skills profileModel Context Protocol (MCP) profileOpen in the compare tool
Side-by-side facts
10 of 16 rows differ (marked). Unknown means no source we track states it.
| Field | Agent Skills | Model Context Protocol (MCP) |
|---|---|---|
| Type | Protocol · specification CUR | Protocol · specification CUR |
| Publisher · differs | Anthropic CUR | Agentic AI Foundation (Linux Foundation) CUR |
| Licence · differs | Apache-2.0 SRC | MIT AND Apache-2.0 CUR |
| Open source | Yes DER | Yes CUR |
| Pricing model | open-source CUR | open-source CUR |
| Hosting · differs | Remote (hosted by the publisher) SRC | Unknown |
| Latest version · differs | 1.0.0 SRC | Unknown |
| Languages · differs | python SRC | typescript SRC |
| GitHub stars · differs | 27k (27,493) SRC | 9.7k (9,737) SRC |
| Forks · differs | 2.1k SRC | 1.9k SRC |
| Last push · differs | 2026-11-04 SRC | 2026-11-21 SRC |
| Archived | No SRC | No SRC |
| First published · differs | 2026-02-04 SRC | Unknown |
| Versions recorded · differs | 1 SRC | Unknown |
| Verification | Checked by AgentSidekick against official sources CUR | Checked by AgentSidekick against official sources CUR |
| Last verified | 2026-11-21 CUR | 2026-11-21 CUR |
Shared and unique capabilitiesDER
Capabilities are assigned by keyword rules from each description unless a researcher set them; integrations are services named in the description or package metadata.
Both (0)
None identified.
Only Agent Skills (0)
None identified.
Only Model Context Protocol (MCP) (1)
Platform compatibility
Status per platform from the evidence we hold. No pairing has been tested by AgentSidekick.
No compatibility evidence
For all 47 platforms on record the status is Unknown or Platform discontinued for both: neither product type implements a standard we can check (MCP, Agent Skills, Claude plugins) and neither publisher names a platform.
Maintenance
Repository activity read from the GitHub API. Activity is not quality.
| Measure | Agent Skills | Model Context Protocol (MCP) |
|---|---|---|
| Status | Active DER | Active DER |
| Days since last push | 17 DER | 0 DER |
| Open issues | 58 SRC | 74 SRC |
| Latest GitHub release | Unknown | 2026-07-28 · 2026-07-28 SRC |
| Newest registry version | 1.0.0 · 2026-02-04 SRC | Unknown |
Evaluation dimensions
Separate signals with their evidence; there is no overall score. Hover or focus a status for its evidence.
| Dimension | Agent Skills | Model Context Protocol (MCP) |
|---|---|---|
| Source authority | Checked against official sources | Checked against official sources |
| Publisher verification | Organisation account | Organisation account |
| Maintenance activity | Active | Active |
| Licence clarity · differs | Apache-2.0 | MIT AND Apache-2.0 |
| Installation clarity · differs | Machine-readable | Linked |
| Documentation · differs | Repository / website only | Docs linked |
| Compatibility evidence | Unknown | Unknown |
| Security review | Not reviewed | Not reviewed |
Which fits which requirementDER
Derived from the differences above, one rule per line. Nothing here says either product is better; check the requirement against your own needs.
Choose Agent Skills if…
- you weigh adoption: 27k vs 9.7k GitHub stars (popularity, not quality)
- you want machine-readable install metadata (packages or endpoints declared in a registry)
Choose Model Context Protocol (MCP) if…
- you need Authorization (classified for Model Context Protocol (MCP) only)
Sources
- Agent Skills · Official MCP Registry: registry.modelcontextprotocol.io (external)checked
- Agent Skills · Curated research: agentskills.io (external)checked
- Agent Skills · Researcher checked: agentskills.io (external)checked
- Agent Skills · Researcher checked: github.com (external)checked
- Model Context Protocol (MCP) · Curated research: modelcontextprotocol.io (external)checked
- Model Context Protocol (MCP) · Researcher checked: modelcontextprotocol.io (external)checked
- Model Context Protocol (MCP) · Researcher checked: modelcontextprotocol.io (external)checked
- Model Context Protocol (MCP) · Researcher checked: github.com (external)checked
Page data last updated .