Field-by-field comparison

Compare the evidence before choosing an MCP.

Each signal stays separate, missing facts remain visible, and the column order follows your selection.

MCPs in this comparison

Uncheck an MCP to remove it. Comparing two or three keeps each signal in its own row.

Open canonical comparison link

Comparable alternatives

Shown only when capability or industry context overlaps.

Side-by-side MCP evidence comparison
Evidence fieldBeadsWatchlist candidateWindows-MCPMCP100 Selected
Task fit
Best for
  • Coding-agent teams that want persistent dependency-aware issue/task memory and specifically need an MCP-only interface.
  • Using an AI client to inspect and operate Windows applications when no safer application API exists
Not ideal for
  • Shell-capable coding agents where the project itself recommends the CLI, or contexts highly sensitive to MCP schema/token overhead.
  • Deterministic automation that can be implemented through a documented API, CLI, or test framework
Avoid when
  • Avoid treating publisher-estimated context overhead as a measured MCP100 Context Footprint; runtime footprint remains unmeasured unless protocol-measured.
  • The desktop session contains sensitive applications, credentials, messages, or files outside the task
Provenance
Provenance details

Ecosystem official

github:gastownhall · Publisher source ↗

Community implementation

io.github.CursorTouch · Publisher source ↗

Maintenance
Maintenance details

Stale

Repo: Not documented

Package: Not documented

Repo: Aug 14, 2026

Package: Aug 1, 2026

Popularity evidence
GitHub starsNot documented

6,749

GitHub stars · checked 2026-08-14T18:46:26.000Z

Package downloads / 30 daysNo reliable download data

40,449

package downloads / 30 days · checked 2026-08-14T18:46:26.000Z

Client coverage
Client coverage detailsNot documented
Claude CodeLocal stdio
Claude DesktopLocal stdio
CodexLocal stdio
VS CodeLocal stdio
CursorLocal stdio
OpenCodeLocal stdio
Client coverage noteLocal stdio runs on your computer. Streamable HTTP connects to a remotely hosted MCP server.
Access and data
AuthenticationTypically local project/workspace access rather than a hosted user OAuth service; exact server authentication is deployment-dependent and must come from bound local evidence.Stdio has no network listener. Non-loopback HTTP requires an auth key or OAuth unless insecure access is explicitly allowed; IP allowlists, TLS, CORS, and host validation are documented.
CostOpen-source project; local database/infrastructure/model costs may apply.The MIT-licensed project lists no MCP fee. Windows, Python, hosting, client, and model costs are separate.
PermissionsCan read/update Beads task/issue state in the configured workspace/database.Runs with the invoking user's full Windows permissions, including PowerShell, files, registry, processes, UI input, screenshots, and web scraping; many actions are irreversible.
Data handlingAgent task descriptions, dependencies and project-management context are persisted in the Beads data store and exposed to the connected client.Processing is local. Telemetry excludes screenshots, state captures, arguments, and outputs but includes status, duration, tool and client information, anonymized sessions, and possibly paths in error messages.
Limitations
Tradeoffs
  • Promoted after strict re-audit: canonical identity and MCP-native/publisher-backed implementation are sufficiently evidenced, with real workflow value, usable documentation, understandable permissions, plausible maintenance, and no unresolved security veto. Still not strong enough for Selected.
  • General computer-use reaches many applications, but visual automation is brittle and broad input control can produce high-impact unintended actions.
Risk contextMCP is a secondary interface and can add material schema/context overhead; protect persistent project/task data.It is not sandboxed and can delete files, alter registry or system settings, type into applications, and execute PowerShell. Use a VM or low-privilege account, backups, supervision, loopback binding, and disabled high-risk tools.
Evidence date
Editorial review2026-09-13T00:00:00Z2026-08-11
Candidate evidence2026-09-13T00:00:00Z2026-08-14T18:46:26.000Z

Popularity, maintenance, fit, permissions, and client support are independent evidence fields.

Read the evidence method →