Model Context Protocol server for Rollbar
Tools
check_all
Check all configured sources now and return ModelWatch results as JSON.
ArgumentsA call is made on an account: it counts against an allowance and the publisher sees it, which is why this one asks who you are first.
99/100
trust score
3
tools declared
remote
via our gateway
—
p95 latency
What it does
ModelWatch checks selected official AI provider sources on demand: Gemini API pricing, Gemini release notes and the latest Anthropic Python SDK release. Call check_all to establish a baseline and compare later observations.
A change remains pending until a second matching observation confirms it. Confirmed changes retain the previous baseline in history. Results link to the official source and can include calendar data for a detected future date. Classification and dates are heuristic: review the source before acting.
Use status to inspect configured sources and baseline presence, and self_test to run the deterministic offline checks. No background polling or automatic alerts. Support: alystheexperiment@gmail.com.
Quickstart
# 1 — buy it first ($9 a month): mcprush.com/checkout?server=modelwatch-v0-3-4-mcp
# a key alone is not enough: the gateway refuses every call until it is bought
# 2 — sign in with a key of the account that bought it: npx mcprush@latest login
# the key is at mcprush.com/dashboard#access
# 3 — install
npx mcprush@latest add ciccio/modelwatch-ai-provider-change-monitor-mcp
# 4 — ask your agent for something these tools do
# check_all, status, self_test
Windows PowerShell says 'running scripts is disabled on this system'? Run Set-ExecutionPolicy -Scope CurrentUser RemoteSigned once, or type npx.cmd instead of npx (codex.cmd, gemini.cmd, grok.cmd likewise). PowerShell 7 needs neither.
PublisherVerified · since 2026
Installation
Where are you running it?
Every route below installs the same thing and ends at the same approval screen. Nothing here runs on your machine — this server runs on infrastructure we host behind our gateway, and what you install is the connection to it.
Claude Codenot declared — it may still work
Claude Code has its own MCP command, and the key rides on it as a header — the gateway refuses a call that carries none. Claude Code reads MCPRUSH_KEY each time it connects, so the key stays out of its config: export your key from Access in your shell profile (macOS, Linux), or run setx MCPRUSH_KEY <your key> and open a new terminal (Windows). Or let npx mcprush@latest add ciccio/modelwatch-ai-provider-change-monitor-mcp write the entry with the key for you; --scope user makes it available in every project. The first connection after a quiet spell starts the server, which takes up to a minute: if your client says it failed to connect, reconnect after that (/mcp in Claude Code).
Buy it first — $9 a month. A key alone is not enough: until this account has bought it, the gateway refuses every call. Subscribe, then use a key of the account that bought it.
claude mcp add --transport http --scope user modelwatch-v0-3-4-mcp https://mcprush.com/gw/modelwatch-v0-3-4-mcp/mcp --header 'Authorization: Bearer ${MCPRUSH_KEY}'Reconnect, or start a new session, and the tools appear in the model’s tool list.
This is a paid listing: the gateway answers only a key of an account whose subscription to it is active — how the login works. Until then add writes nothing: signed out, it names the price and the checkout link; signed in, it prints the checkout link, or why the subscription is held (ended, paused, an unpaid invoice) and the link that lifts it. The gateway refuses every call the same way, and most clients only show the server as failed.
What it writes
One config entry pointing at the gateway. The server itself runs on infrastructure we host, so nothing from this listing executes on your machine.
What you approve
3 tools, with what each one reads, writes and reaches shown before you agree — the same list on every route above. Read the tool surface.
Tool surface
What the model actually sees. Descriptions are diffed on every release — see version history.
Check all configured sources now and return ModelWatch results as JSON.
No parameter schema on file.
Show ModelWatch version, local data paths, configured sources, and baseline presence.
No parameter schema on file.
Run ModelWatch's offline deterministic safety/behavior self-test.
No parameter schema on file.
Standard
$9 / month
- 2,500 calls a month, included
- 30 calls a minute at peak
- Every tool the listing has
- Cancel any time — the unused days are credited
- Runs behind the gateway, so cancelling stops it
What counts against the 2,500 calls a month on Standard
| Tool | Unit | Calls used | Out of the monthly allowance |
|---|---|---|---|
| check_all | write | 2 | 0.0800% |
| status | call | 1 | 0.0400% |
| self_test | call | 1 | 0.0400% |
Failed calls, timeouts and anything the gateway rejects never count. When the allowance runs out the gateway returns an error naming the plan limit — nothing is billed beyond the subscription, ever.
Call allowance · this installenforced at the gateway
Monthly allowance, in calls
—
—
Alert at 80% · then the tool returns a normal MCP error
Used this month: —
ATrust score 99 / 100
Release 0.1.0 · scanned 14 Oct 2026
Two independent axes, because powerful and malicious are different questions. The grade is threat only. The capability level is blast radius, and it is never a penalty on the grade — it is priced as one subtract-only term in the score, where you can see it.
Threat score100
Client score99
Findings0 threat · 0 capability
Capabilityminimal
Coveragesource
Verificationrepo
Surface digestbf12a18b406ef8ce972b3c1e6210f94d270cb5cf5c6dd9c9f0c1e7ce0564a905
Score vectorthreat 100 → client 99
| Term | Level | What it prices | Points |
|---|---|---|---|
| capability-exposure | minimal | capability blast radius (minimal) — client exposure if the model is manipulated | −0 |
| verification-discount | repo | publisher verification (public source) — no provenance, but the source is public and inspectable | −1 |
| coverage-honesty | source | inspection depth (source) — how much of the target the scan could see | −0 |
score = clamp(round(threatScore − E_cap − E_ver − E_cov), 0, 100). No model, no opinion, no network call at scoring time: the same methodology version against the same bytes always produces this number, and every line above is a field of the report rather than a summary of it.
What the scan could actually read
A grade is only as meaningful as its coverage, so the scanner publishes its own depth before it publishes its result.
Tool surfaceread
Implementation sourceread
Package metadataread
Live transportnot available
Tools were statically extracted from the published source (3 recovered), not enumerated from a running server. Tool-poisoning, Unicode-smuggling, capability and toxic-flow analysis ran on this inferred surface, but a mis-parsed registration could be missed or mis-attributed, so tool-derived findings are capped below “confirmed”. To grade the real runtime surface, scan the running server: --command "npx -y <package>".
Capability — what it could do if the model were manipulated
Tags derived from each tool’s schema and the implementation, not from what the tool calls itself. minimal is the level these add up to.
| Tool | Capability tags | Why the tag was assigned |
|---|---|---|
| check_all | no tags | |
| status | no tags | |
| self_test | no tags |
Toxic-flow graph
The lethal trifecta, checked as a graph rather than as a checklist: untrusted input, a sensitive source and an external sink have to meet before there is a path worth worrying about.
The public result for this release does not print the flow graph, so there is nothing to show here. That is not the same as "no paths were found": what the scan did read is above, under coverage.
Supply chain and provenance
Integrityfirst-seen
Surface digestbf12a18b406ef8ce
Scanned2026-10-14
This is the first scan of this surface here, so there is nothing yet to compare it against.
Every result on this tab comes from one deterministic pass over the published package — offline, rule by rule, and auditable line by line above. Same methodology version, same bytes, same score.
Release history
The gateway serves every install the publisher’s current release, so a release cannot be pinned or rolled back to here yet. The history below is for reading; the current release on file is 0.1.0.
0.1.0
8 Oct 2026
latest- The version this listing went live with, after review.
—
no reviews yet
Who can review
One review per account, from a signed-in account that does not publish this listing. Nothing else is asked — you can say what you think before you install it. Publishers can reply once per review, and a review can be edited or deleted by whoever wrote it.
Sign in to write one
Any signed-in account that does not publish ModelWatch — AI Provider Change Monitor can review it, once — before installing it or after. A review from an account that has installed it is marked as one.
No reviews yet
Nobody has reviewed this listing. The rating on the card is the mean of the reviews written here and nothing else, so there is no rating until somebody writes the first — which takes a signed-in account that does not publish it, and nothing else.