Commands
| Command | Meaning |
|---|---|
momonga attach [run-id] | follow a run until it ends; the latest run of this directory when no id is given |
momonga approve [approval-id] [option] | list pending approvals, show one, or resolve it with allow_once, allow_always, reject_once or reject_always |
momonga cancel [run-id] | cancel a run; the running run of this directory when no id is given |
momonga report [run-id] | show the verification report of a run: the declared completion conditions, what passed, failed or was not run, and the residual risk; the latest run of this directory when no id is given |
momonga trace [run-id] | show the model and tool spans of a run; the latest run of this directory when no id is given |
momonga acp | speak the Agent Client Protocol on standard input and output so an editor can drive momonga |
momonga update | replace this executable with the latest release after verifying its checksum and signature |
momonga daemon | manage the background daemon |
momonga daemon start | start the daemon, or start a new one beside it when this program is newer |
momonga daemon status | show whether the daemon runs, its version and the state of its processes |
momonga daemon stop | stop the daemon and the work it runs, after a bounded wait for its processes to end |
momonga gateway | inspect and tune the model gateway |
momonga gateway probe <model> | measure a model of a managed server, an OpenAI-compatible server or an Ollama server and record the result for the gateway (an Ollama server: trained context length, tools, vision, thinking and model digest; a managed or OpenAI-compatible server: tool calling, structured output, stop sequences, parallel calls, reasoning control, usage) |
momonga elicit [input-id] | list pending MCP input requests, show one, or answer it with --accept, --decline, --cancel or --responses |
momonga mcp | inspect and authorize the MCP servers of this folder |
momonga mcp list | list the MCP servers with their transport and authorization state |
momonga mcp login <server> | authorize a remote MCP server in the browser |
momonga auth | sign in to a subscription provider, sign out, and show the saved accounts |
momonga auth login [provider] | sign in to a subscription provider in the browser |
momonga auth logout [account] | sign out a saved account; without a terminal the account is required |
momonga auth status [model] | show the saved accounts, their state, where to manage usage, and the model new sessions start with or the model named |
momonga auth resume <account> | resume plan use of an account paused by a usage limit |
momonga auth reset-host | register this computer as a new host; asks first, or needs --yes without a terminal |
momonga status [run-id] | show a run, the subagent runs under it and the approvals waiting in them; the latest run of this directory when no id is given |
momonga jobs [session-id] | list the background jobs of a session, including those of its subagents; the session of the latest run of this directory when no id is given |
momonga memory | add, review and retire the memories of this folder and of all folders |
momonga memory add <text...> | add a memory as a person; --path binds it to a file of this folder, so it shows as stale once the file changes |
momonga memory list | list the memories |
momonga memory show <id> | show a memory with its source |
momonga memory history <id> | show a memory with the notes it replaced and the ones that replaced it, and their validity |
momonga memory delete <id> | retire a memory: it is no longer recalled or listed and stays in momonga memory history |
momonga memory review | list the memories the agent proposed in this folder and wait for your decision |
momonga memory review accept <id> | accept a proposed memory, so later sessions recall it |
momonga memory review reject <id> | reject a proposed memory, so it is never recalled and never proposed again |
momonga memory restore <id> | bring an archived memory back |
momonga memory promote <id> | write an accepted procedure memory as a skill |
momonga rule | standing rules that bind every run of a folder or of all folders |
momonga rule add <target> <text...> | add a rule; --deny or --ask makes it enforced by the permission system |
momonga rule list | list the rules of this folder and the rules for every folder, enforced or text-only |
momonga rule retire <id> | retire a rule; a retired rule stays in the history |
momonga rule history <id> | show the rules a rule replaced and the ones that replaced it, with their validity |
momonga cost [run-id] | show the model cost of a run, of a session, or of the sessions of this folder |
momonga reliability | report the completion, recovery, duplicate, intervention, budget, time and cost indicators of the terminal tasks of the period |
momonga permissions | review the permission decisions of recorded tool calls |
momonga permissions review [run] | list a sample of recorded tool calls to review with the rates of the reviews stored so far, or store the review of one call |
momonga eval <dataset> | run a golden dataset against the daemon and report the pass rates |
momonga eval context <dataset> | run a context dataset live, each case in its own temporary home with its own daemon, or measure recorded transcripts with --replay, and report the metrics of Evidence-Based Context |
momonga eval yield <dataset> | run a yield dataset live and judge whether reflection pays: each family runs a learn session and then a reuse session with and without the candidates of its reflection, counted as accepted when they hold an anchor of the dataset; exits non-zero when the gate fails |
momonga eval compare <baseline> <candidate> | print the paired differences of two context reports and pass or fail the candidate against the gate |
momonga eval bench <directory> | run every task of the benchmark in the directory on one model, each trial in a fresh copy of its repository, check each with the oracle of its task, and print the reliability report; with --validate, check the benchmark itself without a model |
momonga auto-mode | measure how well a model judges approvals for auto mode, or show the stored measurement |
momonga auto-mode check [model] | run the auto-mode dataset against a classifier model in temporary homes and store the measurement; auto mode stays selectable whatever it shows |
momonga auto-mode status [model] | show the stored measurement of a classifier model for this version |
momonga context | improve how the context is compacted, from the failures a benchmark shows |
momonga context improve | analyse the scenarios that pass without compaction and fail with it, sample candidate guidelines for the compaction summary, benchmark each against the guideline in force, print the winner and its full text, and write the outcome to a file; nothing is stored |
momonga context guideline | inspect, adopt and roll back the compaction guideline |
momonga context guideline adopt <outcome-file> | store the winner of an outcome file that context improve wrote, after checking its digest and its parent and showing its full text |
momonga context guideline list | list the adopted guidelines, oldest first |
momonga context guideline rollback <version> | return to the text of an earlier version, recorded as a new version |
momonga secret | store the secrets that model servers refer to |
momonga secret set <name> | store a secret in the platform secret store; the value is read from the terminal |
momonga models | list the configured models with the capabilities they declare |
momonga models detect [url...] | ask the usual local addresses, and any address you name, what model server answers there; nothing is written |
momonga models check | ask each configured server once whether it still lists its configured models, and list the models with the result |
momonga models catalog [account] | list the models of a signed-in subscription account, read again from the provider; the active account when only one could answer |
momonga models add | add a model of a model server found on this computer, of a server at an address, or of a signed-in subscription account |
momonga models remove [model] | remove a model from the settings, with every setting that names it |
momonga models remove-server [server] | remove a server and its models from the settings, with every setting that names them |
momonga models start [model] | choose the model new sessions start with |
momonga models fallback | the models tried in order when the model cannot answer |
momonga models fallback list | list the fallback models in order |
momonga models fallback add [model] | add a fallback model |
momonga models fallback remove [model] | remove a fallback model |
momonga models fallback move [model] | move a fallback model to another position |
momonga models billing | the billing changes a request may make when it falls back |
momonga models billing list | list the allowed billing changes |
momonga models billing allow [from] [to] | allow requests to fall from one billing kind to another |
momonga models billing forbid [from] [to] | no longer allow requests to fall from one billing kind to another |
momonga models role | the model for summaries, reflection, analysis and the auto-mode classifier |
momonga models role list | list the model of each job |
momonga models role set [job] [model] | choose the model of a job: compaction, reflection, analysis or classifier |
momonga models role clear [job] | let a job use the model of the session |
momonga models set [model] [field] [value] | set a field of a model, for example capabilities.contextWindow |
momonga models unset [model] [field] | clear a field of a model |
momonga models set-server [server] [field] [value] | set a field of a server, for example billing |
momonga models unset-server [server] [field] | clear a field of a server |
momonga models name [server] [name] | name a server |
momonga models add-program | declare a program that MomongaCode starts as a model server, and add the model it serves |
momonga models set-program [id] [field] [value] | change the command, args, readyPath or modelFile of a declared program |
momonga models forget-program [id] | forget a declared program that no server uses |
momonga models timeouts | how long a request waits for the first reply and without any data, in whole minutes, 1 or more |
momonga models effort [level] | the reasoning effort of the models that have none saved for themselves |
momonga models dialects | the declared schema dialects |
momonga models dialects show | print the declared schema dialects as JSON |
momonga models dialects set | replace the declared schema dialects with the object of a JSON file |
momonga policy | check the permission rules of this folder |
momonga policy verify | prove that no allow rule gets around a deny rule, that no rule leaves a route to secrets, and that every rule can take effect; states what is proved and what is not |
momonga trust [folder] | show whether a folder is trusted, with the settings of it waiting for approval, or decide it; the current folder when none is given |
momonga project-check | the command that checks the work of the current folder, kept in the user settings (projectChecks) |
momonga project-check show | print the check of the current folder |
momonga project-check set <command> | make this command the check of the current folder |
momonga project-check clear | remove the check of the current folder |
momonga attach
| Option | Meaning |
|---|---|
--output-format <format> | output format of --print and attach; values: text, json, stream-json; default: "text" |
--after <seq> | show events after this sequence number; default: 0 |
--verbose | print tool activity to stderr |
-h, --help | display help for the command |
momonga report
| Option | Meaning |
|---|---|
--json | print the report as JSON |
-h, --help | display help for the command |
momonga trace
| Option | Meaning |
|---|---|
--json | print the trace as JSON |
--content | print the stored message content of the run that has not expired, instead of the trace; with --json one JSON object per line |
--page-size <count> | with --content: the stored bodies read per request, from 1 to 50; 50 when omitted; default: 50 |
-h, --help | display help for the command |
momonga update
| Option | Meaning |
|---|---|
--check | only report whether a newer release exists |
-h, --help | display help for the command |
momonga daemon status
| Option | Meaning |
|---|---|
--json | print the status as one JSON document |
-h, --help | display help for the command |
momonga gateway probe
| Option | Meaning |
|---|---|
--samples <count> | samples per measured rate |
--step-timeout <seconds> | the bound of each measuring step, in seconds; default: 120 |
--yes | consent to measure a model paid for by use, up to the bound of requests and tokens the measurement can send |
-h, --help | display help for the command |
momonga elicit
| Option | Meaning |
|---|---|
--accept | accept the request |
--decline | decline the request |
--cancel | cancel the request |
--set <name=value> | a field of the form to accept with; repeat for several fields |
--responses <json> | the complete answer as JSON keyed by request key |
-h, --help | display help for the command |
momonga mcp list
| Option | Meaning |
|---|---|
--json | print the servers as JSON |
-h, --help | display help for the command |
momonga mcp login
| Option | Meaning |
|---|---|
--no-browser | print the authorization URL without opening the browser |
-h, --help | display help for the command |
momonga auth login
| Option | Meaning |
|---|---|
--account <label> | the label of the account; the first free label by default |
--no-browser | print the authorization URL without opening the browser |
--reconsent | sign in again to enable plan use for the account |
--yes | save a first sign-in without asking whether the account is the one meant |
-h, --help | display help for the command |
momonga auth status
| Option | Meaning |
|---|---|
--json | print the accounts and the model as JSON |
-h, --help | display help for the command |
momonga auth reset-host
| Option | Meaning |
|---|---|
--yes | reset without asking |
-h, --help | display help for the command |
momonga status
| Option | Meaning |
|---|---|
--json | print the runs and approvals as JSON |
-h, --help | display help for the command |
momonga jobs
| Option | Meaning |
|---|---|
--json | print the jobs as JSON |
-h, --help | display help for the command |
momonga memory add
| Option | Meaning |
|---|---|
--kind <kind> | the kind of memory; values: fact, episode, procedure |
--path <path> | a file of this folder the memory is about; repeat for several |
--user | add a memory for every folder instead of this one |
--json | print the memory as JSON |
-h, --help | display help for the command |
momonga memory list
| Option | Meaning |
|---|---|
--scope <scope> | only memories of this scope; values: user, project |
--project <path> | only memories of this project folder |
--json | print the memories as JSON |
-h, --help | display help for the command |
momonga memory show
| Option | Meaning |
|---|---|
--json | print the memory as JSON |
-h, --help | display help for the command |
momonga memory history
| Option | Meaning |
|---|---|
--json | print the memories as JSON |
-h, --help | display help for the command |
momonga memory review
| Option | Meaning |
|---|---|
--json | print the memories as JSON |
-h, --help | display help for the command |
momonga memory promote
| Option | Meaning |
|---|---|
--skill <name> | the name of the skill to write |
-h, --help | display help for the command |
momonga rule add
| Option | Meaning |
|---|---|
--kind <kind> | the kind of rule; values: constraint, preference, commitment, role, identity; default: "constraint" |
--user | apply in every folder instead of this one |
--path <glob> | apply only when this path pattern is touched; repeat for several |
--deny <rule> | a permission rule to deny, for example "Bash(rm:*)" |
--ask <rule> | a permission rule to ask about, for example "Bash(git push:*)" |
--supersedes <id> | the id of the rule this one replaces; repeat for several |
--until <date> | the date after which the rule no longer applies |
--json | print the result as JSON |
-h, --help | display help for the command |
momonga rule list
| Option | Meaning |
|---|---|
--json | print the rules as JSON |
-h, --help | display help for the command |
momonga rule history
| Option | Meaning |
|---|---|
--json | print the rules as JSON |
-h, --help | display help for the command |
momonga cost
| Option | Meaning |
|---|---|
--session <id> | the cost of one session |
--all | the cost of every session |
--breakdown | show what the requests were made of, the billed tokens by kind, the cache hit ratio, the timing and the cost |
--json | print the report as JSON |
-h, --help | display help for the command |
momonga reliability
| Option | Meaning |
|---|---|
--group-by <key> | the key the indicators are grouped by; values: model, mode, agent, origin, period; default: "model" |
--days <n> | the period: the tasks whose root run ended in the last n days; default: 30 |
--explain | list the run ids behind each count |
--json | print the report as one JSON document |
--out <file> | also store the report as JSON in the file |
--verify | fold the journal of every stored facts row again and list the runs whose stored facts differ |
-h, --help | display help for the command |
momonga permissions review
| Option | Meaning |
|---|---|
--sample [n] | list n calls allowed without a person that have a side effect and n asks; 73 when no number is given |
--seed <seed> | the seed of the sample order |
--call <id> | the call of the run the label is for |
--label <label> | the review of the call: correct, unneeded_ask, missed_ask |
-h, --help | display help for the command |
momonga eval
| Option | Meaning |
|---|---|
--record <directory> | save the transcript of every run in this directory |
--replay <directory> | grade recorded transcripts instead of running the dataset |
-h, --help | display help for the command |
momonga eval context
| Option | Meaning |
|---|---|
--replay <directory> | the recorded transcripts to measure instead of running the dataset |
--record <directory> | save the transcript of every measured run of a live run in this directory |
--model <model> | the model every session of a live run uses |
--seed <n> | the seed of the noise turns of a live run; 0 when omitted |
--report <file> | store the report as JSON for eval compare |
--settings <file> | a token policy overlay over the defaults, every core key on unless the file turns it off (context.adaptiveBudget, context.cachePrefix, context.outputFormatters, context.loopGuidance, context.observedCapacity, context.unboundedOutputReserve, tools.deferral, tools.conciseDescriptions, memory.repositoryFacts, memory.reflection.enabled; context.autoCompact is set by the compaction field of each case and an overlay that sets it is refused), merged into the settings of every case of a live run; the report records the resolved policy |
-h, --help | display help for the command |
momonga eval yield
| Option | Meaning |
|---|---|
--model <model> | the model every session of the run uses |
--report <file> | the file the report is stored in as JSON |
-h, --help | display help for the command |
momonga eval compare
| Option | Meaning |
|---|---|
--gate <name> | the terms the candidate is judged on, one of context, token, reliability: context (the default) and token compare two context reports, reliability compares two reliability reports of one kind (two benchmark reports of eval bench, or two injection reports); default: "context" |
-h, --help | display help for the command |
momonga eval bench
| Option | Meaning |
|---|---|
--validate | check that every task fails on its repository, passes with its reference solution, keeps its grader files out of the repository and names the digest of its repository, and print beside the label of each task the files and lines its reference changes |
--model <model> | the model every task runs on |
--report <file> | store the report as JSON for eval compare --gate reliability |
--group-by <key> | the key the indicators are grouped by; taskType and difficulty are the labels of the declarations; values: model, mode, agent, origin, period, taskType, difficulty; default: "model" |
-h, --help | display help for the command |
momonga auto-mode check
| Option | Meaning |
|---|---|
--agent <model> | the model that drives the runs and proposes the calls the classifier judges |
-h, --help | display help for the command |
momonga context improve
| Option | Meaning |
|---|---|
--dataset <file> | the context dataset to benchmark |
--model <model> | the model of the benchmark sessions and of the analysis and candidate calls |
--token-cap <n> | the most tokens the command may spend on benchmark runs and model calls together; the number of candidates is what this cap affords after the baseline and the analyses |
--seed <n> | the seed of the noise turns and of the candidate draws; 0 when omitted |
--outcome <file> | the file the outcome is written to: the candidates, the text of the winner, its report, the dataset digest, the parent version and a digest of the whole |
-h, --help | display help for the command |
momonga models
| Option | Meaning |
|---|---|
--json | print the models as JSON |
-h, --help | display help for the command |
momonga models detect
| Option | Meaning |
|---|---|
--json | print the servers as JSON |
-h, --help | display help for the command |
momonga models check
| Option | Meaning |
|---|---|
--json | print the models as JSON |
-h, --help | display help for the command |
momonga models catalog
| Option | Meaning |
|---|---|
--json | print the catalog as JSON |
-h, --help | display help for the command |
momonga models add
| Option | Meaning |
|---|---|
--address <url> | the address of the model server; without it this computer is searched |
--model <own name> | the model, as the server calls it |
--window <tokens> | the context window of the model; it wins over the window the server reports |
--api <api> | the API of a server that is not recognised: ollama, openai, anthropic or responses |
--api-key-env <NAME> | the environment variable that holds the key of the server |
--api-key-secret <name> | the secret of momonga secret set that holds the key of the server |
--service <name> | the name of the service, used for prices (for example openai) |
--plan <profile> | add a model of a signed-in subscription account of this provider |
--account <label> | the account of --plan |
--builtin <server> | add the model of a server program built into momonga, for example fm (the Apple on-device model) |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models remove
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models remove-server
| Option | Meaning |
|---|---|
--forget-program | also forget the program the server runs |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models start
| Option | Meaning |
|---|---|
--clear | start with the first model in the settings |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models fallback add
| Option | Meaning |
|---|---|
--position <n> | the position in the order; the end when omitted |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models fallback remove
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models fallback move
| Option | Meaning |
|---|---|
--to <n> | the new position |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models billing allow
| Option | Meaning |
|---|---|
--model <model> | only for this model |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models billing forbid
| Option | Meaning |
|---|---|
--model <model> | only the rule for this model |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models role set
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models role clear
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models set
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models unset
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models set-server
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models unset-server
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models name
| Option | Meaning |
|---|---|
--clear | remove the name, so the server is named by what it is |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models add-program
| Option | Meaning |
|---|---|
--command <path> | the full path of the program |
--arg <text> | an argument of the program; repeat for several, and write {socket} where the program takes the path of its socket |
--ready-path <path> | the path that answers when the program is ready, for example /health |
--name <name> | the name of the server; taken from the program when omitted |
--model <own name> | the model, as the server calls it; read from the server when omitted |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models set-program
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models forget-program
| Option | Meaning |
|---|---|
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models timeouts
| Option | Meaning |
|---|---|
--headers <minutes> | wait for the first reply |
--idle <minutes> | wait without any data |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models effort
| Option | Meaning |
|---|---|
--clear | remove it, so those models run at medium |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga models dialects set
| Option | Meaning |
|---|---|
--file <path> | the JSON file |
--yes | write the change; without it the change is only shown |
-h, --help | display help for the command |
momonga policy verify
| Option | Meaning |
|---|---|
--json | print the report as JSON |
--mode <mode> | the permission mode to check; the starting mode of the settings when omitted; values: manual, acceptEdits, plan, auto, bypassPermissions |
--secret <path> | a further secret location to keep out of reach; repeat for several |
--mandate <name> | check the rules as a run under this mandate of the user settings would have them, and report an allow rule that admits a Bash form a dedicated operation replaces or names an operation that always needs a person |
-h, --help | display help for the command |
momonga trust
| Option | Meaning |
|---|---|
--set <decision> | decide the folder; values: trusted, untrusted |
--clear | return the folder to not decided |
-h, --help | display help for the command |