You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Added
The panel has pages of its own instead of one long Overview: #/sets, #/servers and #/vars are pills in the content head, #/stats and #/set/<id> keep working as before, and every page has an address that
survives a reload and can be sent to someone else.
All servers is a page of the whole host registry: name with the reason of
a failure, transport, a health badge in colour (healthy green, failed red, not checked neutral — the same badge the set table uses), the tools a check
found, the sets the server belongs to, and per-row check, log, edit and × that deletes it from the registry after naming the sets it
leaves. The details drawer no longer carries the registry: it keeps the log
tail of the set, so a set's own composition and a host-wide list stop sharing
one screen.
Variables is a page: the names your servers reference as ${NAME}, each
with whether it is set and where it comes from (openbao, file, env), a
form that writes a value — name, value, store automatic/OpenBao/file —
and × that forgets it everywhere. The panel says plainly that it never
reads a value back and that a name no server mentions stays out of the list.
The panel composes a set in place: the servers table of the Overview tab
gains a registry picker with add to set and a × that detaches a server.
Both send PUT /api/sets/<id> with the whole servers list and redraw from GET /api/state; a refusal shows under the block, an empty set says it will
serve no tools, and the hint states that a running set takes the change by
reload, not by restart.
Import / Export is a dialog of its own, opened from the head of a set and
from the empty state: the whole registry or one set as JSON with Download
JSON and Copy, and an import from a JSON file or a pasted mcpServers
object behind a preview that names the servers to be added — sets found in the
file are skipped with a note, because sets are made in the panel.
The head of a set carries a red Delete set button: the confirmation names
the endpoint and the directory that go away, and the panel returns to the
empty state afterwards.
Prompt in the head of the set opens a dialog over the endpoint below: the
task is required, the introduction and the server subset are optional, and the
built text comes back monospaced with copy and download .md; notes
are shown above it, and a stopped set warns beforehand that schemas will be
missing.
The keys block of a set shows the keyMark of GET /api/state next to the
prefix — the very label the statistics print for that key: its label when it
has one, otherwise key:<mark>; hidden for an open set.
POST /api/sets/<id>/prompt builds one text for an agent out of three
parts: {"task":"…","intro":"…","servers":["…"]} and the answer {"prompt":"…","notes":["…"]}. The tool list comes from the live catalog of
the set — names, descriptions and one-line input schemas — and a stopped set
falls back to the tools.lock.json names with a note that schemas are
missing. One schema is capped at 4000 characters, the whole prompt at 64 KiB,
both cuts reported in notes; set tokens and variable values never reach the
text.
GET /api/state gives every set a keyMark: the label its primary key carries
in the statistics — the key's label when it has one, otherwise the key:<6 hex>
fingerprint — so a row in the statistics can be matched to the key in the
panel. Empty for an open set; keyPrefix stays as it was.
A set may be created and started with no servers. An empty servers list is
now a legal state: POST /api/sets and PUT /api/sets/<id> accept it, in an
update a missing servers key leaves the list untouched, and a set without
servers starts and answers the meta-tools (list_servers is empty). Deleting
the last server of a set no longer answers 400 — the set keeps its key and
port with an empty list.
A saved variable that no server of the install references is visible in the
panel and in GET /api/env with "used": false: the Variables page lists
every stored name, marks the ones nothing references as not used by any
server and lets them be forgotten, instead of hiding a value that stays on
disk. Unreferenced names come last; used is new in the answer.
Docs
Two new diagrams next to the architecture one: the same servers, one
endpoint — what a client pays without a proxy and with one, plus what the
endpoint costs — and what one binary does, a capability map
(docs/diagrams/, spec .json and PNG per language).
docs/en/alternatives.md (and the Russian mirror) compares mcprig with
mcp-proxy, mcp-hub, mcpo, Docker MCP Gateway, MetaMCP, 1mcp, the Lasso gateway
and supergateway, says where each fact came from, and lists where mcprig is
weaker: one operator, no roles, a spoofable X-MCPRig-Caller, a sandbox that
leaves CPU and egress alone.
The README carries a fresh panel screenshot and the corrected count of meta
tools: nine, of which five work with servers.
Changed
A panel started without a password (MCPRIG_AUTH=off) now says so at
startup: WARN панель без пароля (MCPRIG_AUTH=off): доступна всем, кто дотянется до порта; задайте MCPRIG_PASSWORD. mcprig doctor prints the same
as предупреждение: under services when no MCPRIG_PASSWORD or MCPRIG_PASSWORD_HASH is set — a warning, not a problem, so the exit code
stays 0.
list_servers no longer answers with pid: the proxy's process id is of no
use to a client. version stays.
GET /api/calls answers for a stopped set as well: its tail and counters come
from the raw journal on disk (sets/<id>/stats/raw-*.jsonl) over the same
7-day window as period=7d, up to the 512 newest rows, and a missing
directory answers empty instead of landing in unreachable. schema, throttled and unreachable stay properties of a running set; a set without
a token is anonymized on read.
Signing a key (PATCH /api/sets/<id>/keys/<keyID>) reaches the set by reload
instead of a restart: the label travels to the statistics client names and the
set's pid stays put.
Stopping the panel closes every set in parallel under one shared budget (15 s
for readiness and 5 s for termination in total) instead of one budget per set,
so a dozen sets no longer hold the exit for a minute.
A refused start or update now answers by the kind of refusal instead of its
wording: a busy port, a set that did not come up and a set without an active
key answer 409, and anything unclassified answers 500 with a line in the
log and a generic message in the body. A set with an unfilled ${VAR} in its
environment is refused at start with 409 and the name of the variable
instead of starting without it.
mcprig doctor reads an empty install as healthy: a missing config and a
missing registry together mean the setup has not been made yet, so it prints
one line, exits 0 and no longer reports "3 problems" on a fresh directory.
describe_commands answers within the caller's profile: the same allow and
deny rules that hide a tool from list_commands now hide it here as well.
invoke_command takes the server's slot: a slow upstream no longer lets an
authenticated client stack calls past the per-server concurrency limit.
A registry validation error answers 400 with the reason instead of 500
with "see the log": internal/registry now returns a typed ErrInvalid
carrying the text of the first broken field (port 9200, env.A), and the
panel shows it as it is.
A dump at /tool-out/<name> belongs to the key that caused it: the file gets
a <name>.owner fingerprint next to it and another key gets 404. The
lifetime of a dump follows the set policy now (transport.toolOutTtl) instead
of only the panel-wide MCPRIG_TOOLOUT_TTL.
GET /api/servers/<id>/authorize no longer starts the OAuth flow: the flow is POST, and GET …/authorize/status reads it. A GET on the starting route
answers 405 with Allow: POST; the panel polls the status route.
X-Forwarded-Host, X-Forwarded-Proto and X-Forwarded-For are trusted only
from a loopback peer or from an address in MCPRIG_TRUSTED_PROXIES (IP or
CIDR, comma-separated), and not because MCPRIG_PROXY_SECRET is set. A peer
outside the list loses the header silently: the panel answers with its own
origin and the login limit counts the peer's address. A value that parses as
neither an IP nor a CIDR refuses the start with the offending entry.
Behind a trusted proxy the login limit counts the client, not the proxy: five
failed attempts from one X-Forwarded-For address no longer lock out every
other client of the same proxy.
A release is published by CI itself: .goreleaser.yaml carries release.draft: false, so a v* tag produces a public release with no human
step. Every push and pull request runs .gitea/workflows/ci.yml — gofmt, go vet, go build, the race suite, scripts/docs-check.sh, node --check web/app.js and govulncheck — where before only docs were
checked.
Fixed
PUT and DELETE /api/env/<NAME> no longer stop a running set that would
lack a variable afterwards: the set keeps working on its previous environment
and the answer names it in skipped ({set, reason, missing}) next to restarted.
Before, the set was restarted into a refused start and the answer still said
it had been restarted.
Adding a server that brings a ${VAR} the set did not use before restarts the
set: the process environment is fixed at start, so a reload would have left
the variable empty while answering 200. needsRestart compares the variable
references of both compositions, not only the port.
POST /api/sets/<id>/prompt masks tool descriptions and schemas like the
notes, and the mask now includes a server's literal env and headers
secrets on par with ${VAR} values and the set token.
A busy set port is recognised by the typed registry.ErrPortBusy through errors.Is, not by the word «занят» in the error text: rewording the message
can no longer turn a 409 into a 500.
POST /api/browse rejects a server entry carrying a reserved ${VAR} or a {file:…} reference, like every other route that takes a server from the
request body: the entry bypassed the registry, so the browse button could
start a server the set would have refused.
A stdio server now runs in its own process group and its whole group is shut
down when the session closes: SIGTERM, a short wait, then SIGKILL. A server
that wraps its work (sh -c "npx … & wait") left its children behind — the
process was re-parented to init and survived a probe or a set restart, where
only the direct child was signalled. A probe that timed out on initialize
no longer waits out the transport's five-second exit window either: the group
gets the signal at once, so the answer stays inside the probe budget.
A deny rule matches a tool name that spans a newline: the rule pattern is
compiled with . covering every character, so rm* refuses rm\n-rf
instead of letting it through.
The quarantined-tool file is read and written with its errors reported: a
lock file that cannot be read or written refuses invoke_command and invoke_many and logs the reason, instead of quietly turning the quarantine
off.
A probe error printed by list_servers has secrets scrubbed like every other
place that shows upstream text.
A webhook event that could not be delivered is no longer marked as sent: the
next event of the same transition is sent again instead of being suppressed.
GET /api/sets/<id>/servers no longer answers 200 with servers: null when
the probe gave no answer: every server of the set stays in the list with the
probe error, so the panel does not read an expired probe as an empty set. A
client that cancels the request gets no answer; a start or reload cancelled by
the client stops the spawned set process instead of leaving it holding the
port.
A SIGKILL on a set no longer leaves its servers running: every stdio server
starts through a hidden mcprig child-guard -- wrapper that is the leader of
the server's process group, and the kernel hands it SIGTERM when the set
dies, so it kills the whole group before it exits. A server that wraps its
work (sh -c "npx … & wait") was previously re-parented to init and kept the
upstream connection and the disk lock of a set that no longer existed. If the
binary was replaced under a running set, the wrapper cannot be started and the
server runs without it, with a warning in the set log.
A dump created before 0.6.0 has no owner fingerprint, so /tool-out/<name>
answers 404 for it after the upgrade instead of serving the file.
Closing a set no longer waits for another init to reap its killed processes:
a zombie holds no port and no memory, and in a container it stays around, so
the old check spent the whole grace budget on it.
The panel refuses a browser request that carries neither Origin nor Sec-Fetch-Site and a Origin: null or Sec-Fetch-Site: cross-site outright:
before, a request without those headers skipped the origin check, so a
cross-site page could reach an unsafe method over a session cookie. curl
without a cookie keeps working, and /api/login stays outside the check.
A first OAuth connect no longer fails with <server>: context deadline exceeded on a loaded machine: the token cache is written atomically without fsync. Two fsyncs of 300-byte files took up to seconds on a busy disk
(measured 5.3 s and 7.6 s here), and the SDK makes that write inside the
connect whose deadline then expired — the TestOAuthFlowAgainstFakeAS flake
reproduced as 1 failure in 40 runs before the change and 0 in 90 after.
A machine crash may now lose the last token written; the next call
authorizes again.