Skip to content
MCP server

Eclipse Umbra MCP

Watches agent identities for lookalike names and silent endpoint changes

Platforms
API
Version
0.1.0
Listing updated
today

Similar MCP servers

About Eclipse Umbra MCP

Eclipse Umbra MCP is an MCP server for agent identity. It is available on API.

Umbra is aimed at the gap between where an agent identity is published and where it is actually used. It compares listings across registries to find typosquats and near-duplicate names, and it flags cases where an endpoint changes without a visible version change. Findings are limited to a fixed set of states, so a result describes what was observed rather than declaring a party guilty.

Each finding comes with a reproducible evidence pack and is signed under a dedicated findings key, so a reviewer can check it later. The server offers no enforcement actions at any tier: it cannot block, delist, or revoke anything. Affected parties can file a rebuttal against a finding, and the record of that rebuttal is kept with it.

The current version (v0.1) works from snapshots you import yourself. Live crawling of registries is planned but gated per registry and is not built yet. The product is a fit for registry operators, marketplace trust teams, and security staff who want a watchlist of agent names and endpoints with an auditable trail.

Tools this server provides (11)

Taken from the project documentation; the running server may differ.

snapshot_import
Imports a registry snapshot for analysis.
watchlist_set
Sets the identities and endpoints to watch.
diff_run
Compares snapshots to detect endpoint changes.
similarity_scan
Scans for lookalike and typosquat names.
finding_issue
Issues a finding in the closed state vocabulary.
finding_get
Returns one finding with its evidence.
finding_list
Lists issued findings.
finding_verify
Verifies a finding's signature and evidence.
rebuttal_file
Files a rebuttal against a finding.
receipt_list
Lists action receipts.
receipt_verify
Verifies a receipt.

What it does

Lookalike and typosquat scanning
Compares agent names across imported registry snapshots to surface near-identical identities.
Endpoint change detection
Diffs watched endpoints between snapshots and reports silent changes.
Closed finding vocabulary
Findings use a fixed set of states so results stay consistent and comparable.
Signed evidence packs
Each finding ships with reproducible evidence signed under a dedicated key.
Rebuttal filing
Affected parties can file a rebuttal that is recorded alongside the finding.
No enforcement verbs
The server cannot block, remove, or revoke anything; it only reports.

Good for

  • Watch a list of agent identities for impersonation across registries
  • Detect when a published agent endpoint changes between snapshots
  • Build an auditable record of suspected lookalike listings for review
  • Give affected publishers a formal channel to contest a finding

Eclipse Umbra MCP FAQ

Does Umbra crawl registries live?

Not in v0.1. It works from snapshots you import; live crawling is gated per registry and not built.

Can Umbra take action against a flagged identity?

No. The server has no enforcement verb at any tier and only issues findings.

What does a finding include?

A state from a closed vocabulary, a reproducible evidence pack, and a signature under a dedicated findings key.

Can a flagged party respond?

Yes, through the rebuttal_file tool, which records the rebuttal against the finding.

Where this comes from

  • Description: written from the project's website and its MCP Registry entry.

How AIPoweredBot checks servers