High-performance browser automation bridge and multi-instance orchestrator with advanced stealth injection and real-time dashboard.
Config
pinchtab config is the CLI entry point for creating, inspecting, validating, and editing PinchTab's config file.
For security posture, token usage, sensitive endpoint policy, and IDPI guidance, see Security.
Commands
pinchtab config
Prints a read-only config overview. It is not an editor; change values with
config set, config patch, or by editing the file.
It shows these high-signal effective settings:
multiInstance.strategymultiInstance.allocationPolicyinstanceDefaults.stealthLevelinstanceDefaults.tabEvictionPolicyinstanceDefaults.tabPolicy.lifecycle(with the close delay when an idle lifecycle is on)
It also shows:
- the active config file path
- the masked server token
- the dashboard URL when the server is running, otherwise
not running - hints for
config get,config set,config show,config tokenandpinchtab security
pinchtab config
pinchtab config init
Creates a default config file at the current config path.
pinchtab config init
config init respects PINCHTAB_CONFIG. If that environment variable is set, the file is created there.
Generated config files include a $schema URL for IDE completion and validation.
The new file gets a generated server.token, reported on stderr. If a file already
exists at that path, config init asks before overwriting it.
pinchtab config schema
Prints the JSON Schema URL for this PinchTab build. Source builds, development
builds, and versions without a published schema use the main schema URL.
For a release build PinchTab uses the closest published schema tag at or below its
own version, so a generated config never points at rules from a newer release; with
no such tag it uses main.
pinchtab config schema
Print the bundled schema JSON:
pinchtab config schema --print
pinchtab config show
Shows the effective runtime configuration.
pinchtab config show
Secret values such as server.token remain masked in this output.
The Security section also includes security.trustLoopbackProxy as Trust Loopback Proxy so proxy trust posture is explicit.
pinchtab config token
Copies the configured server.token to the system clipboard. Nothing is written
to stdout, so a shell capture returns empty rather than a message.
pinchtab config token # copy to clipboard; status goes to stderr
TOKEN=$(pinchtab config token --stdout) # print the token and nothing else
--stdout is the supported way to read the token on a host with no clipboard
tool. Without it the token is never printed.
If clipboard access is unavailable the command exits non-zero and names both
--stdout and the config file path on stderr, so a script sees the failure
instead of a success that copied nothing.
pinchtab config path
Prints the config file path PinchTab will read.
pinchtab config path
pinchtab config validate
Validates the current config file.
pinchtab config validate
pinchtab config get
Reads a single dotted-path value and reports the value in effect: what the file
sets, and otherwise the value the runtime resolves — a shipped default, or a value
derived from another key. profiles.baseDir is the clearest case: leave it out of the
file and config get reports <server.stateDir>/profiles, the directory profiles are
actually kept in, rather than nothing. A key that is genuinely unset — a secret nobody
configured, an optional override such as browser.binary — still answers blank.
pinchtab config get server.port
pinchtab config get instanceDefaults.mode
pinchtab config get security.attach.allowHosts
pinchtab config get profiles.baseDir # derived from server.stateDir when unset
pinchtab config set
Sets a single dotted-path value in the file config.
pinchtab config set server.port 8080
pinchtab config set instanceDefaults.mode headed
pinchtab config set multiInstance.strategy explicit
pinchtab config patch
Merges a JSON object into the config file.
pinchtab config patch '{"server":{"port":"8080"}}'
pinchtab config patch '{"instanceDefaults":{"mode":"headed","maxTabs":50}}'
pinchtab config patch '{"observability":{"activity":{"retentionDays":14}}}'
Load Order
PinchTab applies configuration in this order:
- built-in defaults
- the config file selected by
PINCHTAB_CONFIGor the default path PINCHTAB_TOKEN, if set, overridingserver.tokenat runtime
Supported environment variables:
PINCHTAB_CONFIG: choose the config file pathPINCHTAB_TOKEN: override the API token at runtimePINCHTAB_RATE_LIMIT_MAX: per-client request cap per 10-second window (default 3000, sized for agent-driven snapshot/action bursts). Lower it (e.g. to 300) when exposing the port beyond localhost. Child instances inherit it from the orchestrator's environment.PINCHTAB_STATE_KEY: state-file encryption key; when set it wins oversecurity.stateEncryptionKey
For remote CLI targeting, use the root --server flag instead of config.
Config File Location
Default location by OS:
- macOS:
~/.pinchtab/config.json - Linux:
~/.pinchtab/config.json - Windows:
%APPDATA%\pinchtab\config.json
On macOS and Linux, PinchTab defaults to ~/.pinchtab so the CLI, npm-managed binary, and config file all use the same base directory.
If you are upgrading from an older macOS setup and still have a config at ~/Library/Application Support/pinchtab/config.json, treat that as a legacy location and move or merge it into ~/.pinchtab/config.json.
Override the config path with:
export PINCHTAB_CONFIG=/path/to/config.json
Config Shape
Current nested file-config shape:
{
"$schema": "https://raw.githubusercontent.com/pinchtab/pinchtab/main/schema/config.json",
"configVersion": "0.8.0",
"server": {
"port": "9867",
"bind": "127.0.0.1",
"token": "your-secret-token",
"stateDir": "/path/to/state",
"logLevel": "info",
"networkBufferSize": 100,
"retainNetworkBodies": false,
"retainNetworkBodyMaxBytes": 262144,
"trustProxyHeaders": false,
"cookieSecure": null
},
"browser": {
"version": "144.0.7559.133",
"binary": "/path/to/chrome",
"remoteDebuggingPort": null,
"extraFlags": "--disable-gpu",
"cloak": {
"fingerprintSeed": "42069",
"platform": "windows",
"locale": "en-GB",
"timezone": "Europe/London",
"webrtcIP": "auto",
"fontsDir": "/path/to/fonts",
"storageQuotaMB": 2048,
"disableDefaultStealthArgs": true
},
"extensionPaths": ["/path/to/pinchtab/extensions"]
},
"browsers": {
"default": "chrome"
},
"instanceDefaults": {
"mode": "headless",
"noRestore": false,
"timezone": "Europe/Rome",
"blockImages": false,
"blockMedia": false,
"blockAds": false,
"maxTabs": 20,
"maxParallelTabs": 0,
"userAgent": "",
"noAnimations": false,
"captureAllowActivation": true,
"humanize": false,
"stealthLevel": "light",
"tabEvictionPolicy": "close_lru",
"tabPolicy": {
"lifecycle": "keep",
"closeDelaySec": 300,
"restore": false
},
"dialogAutoAccept": false
},
"security": {
"allowEvaluate": false,
"allowMacro": false,
"allowScreencast": false,
"allowDownload": false,
"allowCookies": false,
"allowNetworkIntercept": false,
"allowMemory": false,
"allowFileScheme": false,
"allowedDomains": ["127.0.0.1", "localhost", "::1"],
"downloadAllowedDomains": [],
"downloadMaxBytes": 20971520,
"memorySnapshotMaxBytes": 536870912,
"allowUpload": false,
"allowClipboard": false,
"allowStateExport": false,
"stateEncryptionKey": null,
"uploadMaxRequestBytes": 10485760,
"uploadMaxFiles": 8,
"uploadMaxFileBytes": 5242880,
"uploadMaxTotalBytes": 10485760,
"maxRedirects": -1,
"trustedProxyCIDRs": [],
"trustedResolveCIDRs": [],
"trustLoopbackProxy": false,
"attach": {
"enabled": false,
"allowHosts": ["127.0.0.1", "localhost", "::1"],
"allowSchemes": ["ws", "wss", "http", "https"],
"forwardProxyAuth": false
},
"idpi": {
"enabled": true,
"strictMode": true,
"scanContent": true,
"wrapContent": true,
"customPatterns": [],
"scanTimeoutSec": 5,
"shieldThreshold": 30
}
},
"profiles": {
"baseDir": "/path/to/profiles",
"defaultProfile": "default",
"quarantineKeep": 1
},
"multiInstance": {
"strategy": "always-on",
"allocationPolicy": "fcfs",
"instancePortStart": 9868,
"instancePortEnd": 9968,
"restart": {
"maxRestarts": 20,
"initBackoffSec": 2,
"maxBackoffSec": 60,
"stableAfterSec": 300
}
},
"timeouts": {
"actionSec": 30,
"navigateSec": 60,
"shutdownSec": 10,
"waitNavMs": 1000
},
"autoSolver": {
"enabled": false,
"autoTrigger": true,
"triggerOnNavigate": true,
"triggerOnAction": true,
"maxAttempts": 8,
"solverTimeoutSec": 30,
"retryBaseDelayMs": 500,
"retryMaxDelayMs": 10000,
"solvers": ["cloudflare", "semantic"],
"llmProvider": "",
"llmFallback": false,
"external": {
"capsolverKey": "",
"twoCaptchaKey": ""
}
},
"scheduler": {
"enabled": false,
"strategy": "fair-fifo",
"maxQueueSize": 1000,
"maxPerAgent": 100,
"maxInflight": 20,
"maxPerAgentInflight": 10,
"resultTTLSec": 300,
"workerCount": 4,
"maxBatchSize": 50
},
"observability": {
"activity": {
"enabled": true,
"sessionIdleSec": 1800,
"retentionDays": 30,
"events": {
"dashboard": false,
"server": false,
"bridge": false,
"orchestrator": false,
"scheduler": false,
"mcp": false,
"other": false
}
}
},
"sessions": {
"dashboard": {
"persist": true,
"idleTimeoutSec": 604800,
"maxLifetimeSec": 604800,
"elevationWindowSec": 900,
"persistElevationAcrossRestart": false,
"requireElevation": false
},
"agent": {
"enabled": true,
"mode": "preferred",
"idleTimeoutSec": 1800,
"maxLifetimeSec": 86400
}
}
}
sessions.agent.* is described in Agent Identity;
sessions.agent.mode accepts off or preferred. browser.proxy,
browser.targets, browser.defaultTarget and browser.fallbackOrder are covered
under Browser Selection below.
autoSolver.external is config-file-only. Capsolver and 2Captcha credentials
are stored there.
Semantic Flow Credentials
The semantic-first autosolver flow injects credential values into recognised
login/signup/form fields. Configure them under autoSolver.credentials in
the config file:
{
"autoSolver": {
"credentials": {
"login": { "user": "[email protected]", "password": "..." },
"signup": { "name": "Jane Doe", "email": "[email protected]", "password": "..." },
"form": { "field1": "...", "field2": "...", "email": "[email protected]" }
}
}
}
Notes:
- Edit credentials by writing the config file directly. The dashboard config API redacts them on read (GET returns blanks) and preserves on-disk values when a PUT comes in with empty fields, so secrets never round-trip through the UI.
- The form solver step 2 falls back to
form.emailwhenform.field2is empty. - Steps without a configured value fall through to a click-only flow (e.g. a login attempt with no password becomes a "click submit" attempt).
The dashboard Settings page exposes the non-secret AutoSolver settings and shows the active config file path. Provider keys remain managed directly in the config file.
Browser Selection
The CLI uses --browser <name> to select a browser. In the config file the
equivalent field is browsers.default:
{
"browsers": { "default": "cloak" }
}
browsers.default explicitly selects the local browser backend:
chromeuses the normal Chrome/Chromium launch path.ghost-chromeserves static-friendly reads from a lightweight fetcher and escalates to Chrome when a page needs rendering.cloakuses a discovered or configured local CloakBrowser Chromium binary.
When browsers.default is omitted, PinchTab prefers a discovered local
CloakBrowser installation and otherwise falls back to Chrome. Newly generated
config files make the same choice, while existing configs that explicitly set
browsers.default to chrome continue to use Chrome.
browsers.available optionally restricts which browsers requests may select.
For multiple named configurations of the same provider (different binaries,
proxies, or fingerprints), use browser.targets with browser.defaultTarget
and browser.fallbackOrder — see Terminology.
The legacy browser.provider field is no longer supported and is rejected at
validation time.
When the selected browser is cloak, PinchTab uses browser.binary when set
or searches its normal local CloakBrowser paths. A named CloakBrowser target
must set that target's binary explicitly. PinchTab does not download, bundle,
or redistribute the CloakBrowser binary.
On macOS, prefer a dedicated automation browser over your daily Google
Chrome. Launching /Applications/Google Chrome.app headless makes macOS treat
Chrome as already running, so opening your normal Chrome from the Dock just
activates the windowless automation process and no window appears (issue #583).
PinchTab's discovery now prefers Google Chrome for Testing, Chromium, then Chrome
Canary, and only falls back to your primary Chrome as a last resort. If only your
daily Chrome is installed, install Google Chrome for
Testing or set
browser.binary to a separate Chrome/Chromium build. pinchtab doctor browsers
warns when automation would use your primary Chrome.
browser.cloak maps supported CloakBrowser fingerprint settings to native launch
flags:
fingerprintSeed->--fingerprintplatform->--fingerprint-platformlocale->--fingerprint-localetimezone->--fingerprint-timezonewebrtcIP->--fingerprint-webrtc-ipfontsDir->--fingerprint-fonts-dirstorageQuotaMB->--fingerprint-storage-quota
disableDefaultStealthArgs defaults to true for CloakBrowser targets. When set,
PinchTab keeps its process, profile, tab, extension, and action-control behavior,
but does not add its own JS stealth overlays or automation-hiding launch flags.
Set it to false only when you intentionally want PinchTab's legacy stealth layer
on top of CloakBrowser's native patches.
Advanced CloakBrowser flags can still go through browser.extraFlags when they
are not PinchTab-owned lifecycle flags.
browser.proxy.server is the prerequisite for the rest of the block:
browser.proxy.username, password, bypassList and every browser.proxy.geo.*
key do nothing without it, because there is no proxy to route through,
authenticate to, or align geo with. The values are still kept — config set
writes them, config get returns them, and setting the server afterwards makes
them take effect — and PinchTab reports the incomplete block on each config read
or write until a server is set. Clearing browser.proxy.server turns the proxy
off and leaves the other values on disk for the next server.
browser.proxy.username and password are the exception, and it is not about
the server: each requires the other, so config set aborts on either one alone
whichever order you try. Send the pair in one write instead:
pinchtab config patch '{"browser":{"proxy":{"username":"bob","password":"s3cret"}}}'
browser.proxy.geo is a CloakBrowser fingerprint-alignment hint. When a proxy
server and geo block are configured for a CloakBrowser target, PinchTab maps the
geo values into native CloakBrowser fingerprint flags unless the target already
sets the corresponding browser.cloak field. The stock chrome browser does
not derive --lang, TZ, or WebRTC launch settings from proxy geo data.
Browser Extra Flags
browser.extraFlags is validated and sanitized. It is only for user-safe Chrome flags that do not weaken browser security and do not override PinchTab-owned launch behavior.
Rejected examples include:
--no-sandbox--disable-web-security--ignore-certificate-errors--user-agent=...--enable-automation=...--disable-blink-features=...
Use the dedicated config fields instead:
instanceDefaults.userAgentfor UA overridesinstanceDefaults.modefor headed/headlessinstanceDefaults.timezonefor timezonebrowser.extensionPathsfor extension loadingbrowser.remoteDebuggingPortfor the remote debugging port
For Linux container compatibility, use the runtime-managed path instead of browser.extraFlags. PinchTab enables --no-sandbox automatically when needed.
By default, PinchTab looks for unpacked Chrome extensions in <server.stateDir>/extensions. On a normal local install that means the OS-specific PinchTab config directory plus extensions/, for example:
- macOS:
~/.pinchtab/extensions - Linux:
~/.pinchtab/extensions - Windows:
%APPDATA%\\pinchtab\\extensions
You can change or clear that default with browser.extensionPaths.
Tab Policy
instanceDefaults.tabPolicy groups tab lifecycle behavior:
{
"instanceDefaults": {
"tabPolicy": {
"eviction": "close_lru",
"lifecycle": "keep",
"closeDelaySec": 300,
"restore": false
}
}
}
evictioncontrols what happens whenmaxTabsis reached:close_lru,close_oldest, orreject.lifecyclecontrols idle lifecycle behavior:keepdisables lifecycle auto-close and is the default;close_idleauto-closes a tab after it handles an authorized/text,/snapshot, or/actionrequest;freeze_idleinstead freezes a tab no request has touched for the idle delay (timers and JavaScript stop, the page and session stay); every request restarts that clock and unfreezes the tab first. If the renderer does not accept the unfreeze, the request answers503with codetab_unfreeze_failedandretryable: true; the tab and a session's current-tab pointer are kept, and the next request retries. A tab is never frozen while a request on it is running (including a screencast stream), while it is paused for handoff, or while it holds network interception rules.closeDelaySecis the idle delay forclose_idleandfreeze_idle. The default is300seconds when either is enabled.restorecontrols whether session tabs are restored on startup. The default isfalse.
instanceDefaults.tabEvictionPolicy is still accepted for compatibility. New configs should use instanceDefaults.tabPolicy.eviction.
Humanized Input
instanceDefaults.humanize controls whether click and typing actions use the slower humanized path by default. The default is false, which keeps automation fast and deterministic by using raw CDP input.
Callers can override the instance default per action with the JSON field humanize:
{"kind":"click","selector":"#submit","humanize":true}opts a single action into bezier mouse movement and human-like delays.{"kind":"type","selector":"#name","text":"Ada","humanize":true}opts a single type action into the slower per-character path.
Rationale: humanized input is useful for compatibility with pages that react poorly to raw input, but it adds sleeps and multi-step pointer movement. Keeping it opt-in prevents accidental seconds of overhead in default E2E and agent runs.
Sections
| Section | Purpose |
|---|---|
server |
HTTP server settings, log level, proxy trust, cookie transport, and network buffer defaults |
browser |
Chrome executable, version pin, extra flags, extension paths, CloakBrowser flags, proxy, and named targets |
browsers |
Default browser selection and the optional available allowlist |
instanceDefaults |
Default behavior for managed instances |
security |
Sensitive feature gates, transfer limits, attach policy, and IDPI |
profiles |
Profile storage defaults |
multiInstance |
Orchestrator strategy, allocation, port range, and restart policy |
timeouts |
Action, navigation, shutdown, and navigation wait delays |
scheduler |
Optional task queue |
observability |
Activity logging, source selection, and retention |
sessions |
Dashboard session cookies and agent sessions |
autoSolver |
Challenge auto-solver behavior, provider keys, and credentials |
config get And config set Support
pinchtab config get and pinchtab config set accept a section.field dotted path in
these top-level sections:
serverbrowserbrowsersinstanceDefaultssecurityprofilesmultiInstancetimeoutsschedulerobservabilitysessionsautoSolver
Every leaf in those sections is reachable, with these exceptions:
server.engineandbrowser.providerare removed settings; setting them is refusedinstanceDefaults.headlessis superseded byinstanceDefaults.modebrowsers.config.*is a retired block, superseded bybrowser.targetsobservability.activity.stateDircan be read but not set (see Activity Retention)
List values such as security.allowedDomains or browser.extensionPaths are set as a
comma-separated string. $schema and configVersion are document metadata, not settings.
Common Examples
Headed Mode
{
"instanceDefaults": {
"mode": "headed"
}
}
Log Level For A Daemon Or Auto-Started Server
pinchtab server --log-level only reaches a server you start by hand. A
daemon-installed server (pinchtab daemon install) and the server a bare
pinchtab nav auto-starts both launch pinchtab server with no flags, so
server.logLevel is the only way to set their threshold — and it needs no change
to the unit file.
pinchtab config set server.logLevel warn # Warnings and errors only
pinchtab config set server.logLevel debug # Full debug detail while diagnosing
pinchtab config get server.logLevel
Accepted values are debug, info (the default), warn and error; an
unparseable value fails when the config loads, naming the accepted values.
--log-level still wins over the configured value. -v keeps its startup banner
either way, but it only raises the level when neither --log-level nor
server.logLevel is set — so a persisted warn survives pinchtab server -v, and
--log-level debug is how you override it for one run. pinchtab bridge reads the
same key and accepts the same flag.
Memory Diagnostics
| Setting | Default | Effect |
|---|---|---|
security.allowMemory |
false |
Enables POST /memory/snapshot, GET /memory/snapshot/{snapshotId}/summary and GET /memory/compare (code memory_disabled when off). GET /memory needs no capability. |
security.memorySnapshotMaxBytes |
536870912 (512 MB) |
Size cap for one heap snapshot, max 4 GB; a larger snapshot is aborted with memory_snapshot_too_large and no file is left behind |
A heap snapshot contains every string on the page, tokens included, which is why it sits behind its own capability. See memory.md.
pinchtab config set security.allowMemory true
Network Bind With Token
pinchtab config set server.bind 0.0.0.0
pinchtab config set server.token secret
pinchtab server
Changing server.bind away from loopback is a documented, non-default, security-reducing deployment change. Use it only when remote reachability is intentional, keep a token set, and review the outer network boundary explicitly.
If the dashboard is served over plain HTTP on a non-loopback bind, PinchTab shows an in-product warning because session cookies are no longer transport encrypted. Prefer HTTPS or localhost when possible.
Dashboard Cookie Transport
server.cookieSecure controls whether the dashboard session cookie must use the
Secure flag:
null/ unset /auto: default behavior. Session cookies areSecureon HTTPS and non-Secureon plain HTTP.true: always requireSecure. Dashboard login works only over HTTPS.false: always omitSecure, even on HTTPS. Use only for operator-managed edge cases.
Examples:
pinchtab config set server.cookieSecure true
pinchtab config set server.cookieSecure false
pinchtab config set server.cookieSecure auto
When server.cookieSecure = true, plain-HTTP dashboard login fails explicitly
with an HTTPS-required error instead of appearing to succeed and looping.
If TLS terminates in front of PinchTab, also set server.trustProxyHeaders=true
only when the proxy is trusted and rewrites Forwarded / X-Forwarded-*
headers correctly.
Custom Instance Port Range
{
"multiInstance": {
"instancePortStart": 8100,
"instancePortEnd": 8200
}
}
Attach Policy
{
"security": {
"attach": {
"enabled": true,
"allowHosts": ["127.0.0.1", "localhost", "chrome.internal"],
"allowSchemes": ["ws", "wss", "http", "https"],
"forwardProxyAuth": false
}
}
}
security.attach.allowHosts is an allowlist. If you set it to ["*"], PinchTab accepts any reachable attach host with an allowed scheme. That is a documented, non-default, security-reducing override: it removes host allowlisting entirely and should only be used on isolated, operator-controlled networks.
security.attach.forwardProxyAuth controls whether PinchTab may send configured proxy authentication credentials over remote CDP attach. It defaults to false; enable it only when the attached browser process and CDP transport are trusted.
Quarantined Profile Retention
When a browser profile turns out to be unusable, PinchTab renames it aside as
<profile>.quarantine-<unix> and starts fresh. PinchTab keeps the most recent
quarantined copy of each profile and prunes the older ones at the moment a new quarantine
is created — the newest is the copy most likely to relate to a problem being investigated
now, and nothing in the product reads the older ones. Every removal is logged with the
path and the bytes reclaimed.
{
"profiles": {
"quarantineKeep": 1
}
}
profiles.quarantineKeep defaults to 1. Set it to 0 to keep every quarantined
profile, which is exactly the behaviour before this setting existed. Pruning only ever
runs when a new quarantine is created, never on startup or a timer, so quarantined copies
of a profile that never fails again are left alone — clearing that backlog is a separate,
explicit operation.
Activity Retention
{
"observability": {
"activity": {
"retentionDays": 14,
"sessionIdleSec": 1800
}
}
}
Activity logs are always written to <server.stateDir>/activity, so two instances cannot
end up sharing one log directory. observability.activity.stateDir is therefore not
settable: config set refuses it, and config get observability.activity.stateDir reports
the derived directory in effect. Move the logs with server.stateDir.
A file that already carries the key keeps loading and keeps working. The key is ignored,
and saying so is an advisory, not a validation error: PinchTab reports it — at load, and
on stderr when you run config set, config patch or config validate — and it never
blocks anything. There is nothing to fix and nothing to remove; the value simply has no
effect. Validation errors, such as an out-of-range server.port, still block a save.
server.trustProxyHeaders should stay false unless PinchTab is behind a trusted reverse proxy that overwrites Forwarded and X-Forwarded-* headers. Do not enable it on direct-exposure deployments or behind proxies that pass client-supplied forwarding headers through unchanged. When it is enabled, the client-most forwarded address is also the client identity: rate-limit buckets, the concurrent-stream cap and every audit line are keyed on it rather than on the proxy's address.
Legacy Flat Format
Older flat config is still accepted for backward compatibility:
{
"port": "9867",
"headless": true,
"maxTabs": 20,
"allowEvaluate": false,
"timeoutSec": 30,
"navigateSec": 60
}
Use pinchtab config init to create the current nested format.
Validation
pinchtab config validate checks, among other things:
- valid
instanceDefaults.mode - valid
instanceDefaults.stealthLevel - valid
instanceDefaults.tabEvictionPolicy - valid
instanceDefaults.tabPolicy.eviction - valid
instanceDefaults.tabPolicy.lifecycle - non-negative
instanceDefaults.tabPolicy.closeDelaySec instanceDefaults.maxTabs >= 1instanceDefaults.maxParallelTabs >= 0- valid
multiInstance.strategy - valid
multiInstance.allocationPolicy - valid
multiInstance.restart.*values - valid
security.attach.allowSchemes multiInstance.instancePortStart <= multiInstance.instancePortEndmultiInstance.restart.initBackoffSec <= multiInstance.restart.maxBackoffSec- non-negative timeout values
server.networkBufferSizebetween 1 and 10000server.retainNetworkBodyMaxBytesbetween 0 and 10 MiBsecurity.downloadMaxBytes,memorySnapshotMaxBytesand theupload*limits between 1 and their caps, withuploadMaxFileBytes <= uploadMaxTotalBytes- non-negative
security.idpi.scanTimeoutSec - non-negative
observability.activity.sessionIdleSecand positiveretentionDays - valid
sessions.agent.modeand positivesessions.dashboard.*Secvalues server.engineandbrowser.providerare rejected as removed settings
Valid enum values:
| Field | Values |
|---|---|
instanceDefaults.mode |
headless, headed |
instanceDefaults.stealthLevel |
light, medium, full |
instanceDefaults.tabEvictionPolicy |
reject, close_oldest, close_lru |
instanceDefaults.tabPolicy.eviction |
reject, close_oldest, close_lru |
instanceDefaults.tabPolicy.lifecycle |
keep, close_idle, freeze_idle |
multiInstance.strategy |
simple, explicit, simple-autorestart, always-on, no-instance |
multiInstance.allocationPolicy |
fcfs, round_robin, random |
security.attach.allowSchemes |
ws, wss, http, https |
security.attach.forwardProxyAuth |
true, false |
sessions.agent.mode |
off, preferred |
Notes
config showreports effective runtime values, not just raw file contents.config getreports the value in effect, including shipped defaults and values derived from another key.config setandpatchwrite the file config model and do not carry transient runtime overrides.- the dashboard config API treats
server.tokenas write-only; use the CLI or file editing to manage it.