Skip to content

Reference

Plugin registries

A registry is a fixed list of names that briar accepts in one place. This page lists every one, with what each name is and where you use it.

What happens with a wrong name

When you pass a name that is not on the list, briar stops and prints the names it knows. Two env var lists are softer: an unknown name in BRIAR_NOTIFY_SINKS or BRIAR_JOURNAL_SINKS is skipped. These lists match briar-cli 1.1.75. For your own version, run the command with -h: it prints the allowed names.

Extract

An extractor reads one kind of fact from a service and writes it as a section of a knowledge blob. Extractors get their data through the providers further down. Most read your code host. active-tickets and ticket-archaeology read your tracker, aws-infra reads your cloud, and meeting-digest reads your meeting tool.

Scheduled extractors

Where you use it: briar extract --include NAME (you can repeat it), and the extract: list of a runbook schedule. In the source: briar.extract.EXTRACTORS.

NameWhat it is
active-ticketsOpen tickets in your tracker projects.
active-workOpen PRs across your repos.
aws-infraYour cloud setup: servers, databases, queues and logs, on AWS, Google Cloud or Azure.
ci-healthCI pass rate, flaky workflows, and how run time changes.
code-hotspotsFiles that often change together, so the agent knows what else to look at.
code-scanningOpen code scanning alerts, by rule and by file.
codebase-conventionsEach repo's language, test runner, linter and migration tool.
commit-message-qualityHow well commit messages follow Conventional Commits, and how clean their subject lines are.
defect-hotspotsFiles most likely to break, scored from churn, bug fixes and size.
dependency-healthOpen security alerts on your packages, by severity.
github-deploymentsEnvironments, deployments and recent CI runs.
meeting-digestRecent meetings: a summary and the action items of each.
pr-archaeologyPatterns in merged PRs: what reviewers look at, and who reviews.
pr-hygienePR sizes, the share of big PRs, approvals with no comments, and time to first review.
release-cadenceHow often a repo ships, and when it last did.
repo-governanceBranch rules, and whether CODEOWNERS, pre-commit and linter or editor config exist.
revert-signalsReverts and emergency fixes: weak spots that tests and review missed.
review-nitsPhrases reviewers repeat across PRs. Good candidates for lint rules.
reviewer-profileFor each reviewer: how often they comment, on which files, and what they ask for.
stale-prsOpen PRs with no activity for a while: the review backlog.
test-disciplineTest files per source file, and source files with no clear test.
ticket-archaeologyPatterns in closed tickets: who takes them, and which labels, over time.
todo-densityHow many TODO, FIXME and HACK notes there are, and which files have the most.

Task extractors

Where you use it: You do not name these. briar agent (and plan run, which calls it) runs them for the one ticket or PR it works on. In the source: briar.extract.TASK_SCOPED_EXTRACTORS.

NameWhat it is
ticket-contextThe full text, acceptance criteria and comments of one ticket. Used by agent implement.
pr-review-contextThe review comments and CI results of one PR. Used by agent prfix.
meeting-contextMeeting transcripts about the task. Runs when meeting keys are set, or with --meeting-key or --meeting-query.
slack-contextSlack threads about the task, read only. Runs when Slack keys are set, with --slack-query or the ticket key as the search.

Code hosts

Where you use it: --provider on extract, agent and plan run, and provider under [repo] in .briar.toml. Default github. In the source: briar.extract._providers.PROVIDERS.

NameWhat it is
githubGitHub. Reads GITHUB_TOKEN.
bitbucketBitbucket Cloud. Reads BITBUCKET_ACME_USERNAME, _APP_PASSWORD and _WORKSPACE. In 1.1.75 the agent pushes its branch here but cannot open the PR.

Trackers

Where you use it: --tracker on extract, agent implement and plan run, and tracker in .briar.toml. In the source: briar.extract._trackers.TRACKERS.

NameWhat it is
jiraJira Cloud, by API token or browser session. Reads JIRA_ACME_*.
github-issuesGitHub Issues. Reads GITHUB_TOKEN.
bitbucket-issuesBitbucket Cloud issues. Reads BITBUCKET_ACME_*.
linearLinear. Reads LINEAR_ACME_TOKEN.

Clouds

Where you use it: --cloud on extract, for the aws-infra extractor. Default aws. In the source: briar.extract._clouds.CLOUDS.

NameWhat it is
awsAmazon Web Services. Built in.
gcpGoogle Cloud. Needs pip install "briar-cli[gcp]".
azureMicrosoft Azure. Needs pip install "briar-cli[azure]".

AWS services

Where you use it: --aws-extract-service NAME on extract (you can repeat it). Default: all of them. In the source: briar.extract.aws_services.AWS_SERVICE_GATHERERS.

NameWhat it is
ecsECS clusters and their services.
rdsRDS database instances.
lambdaLambda functions.
sqsSQS queues.
logsCloudWatch log groups, the 10 biggest.
tagging-inventoryEvery tagged resource in the region, of any service.

Meeting tools

Where you use it: --meeting on extract. Agent and plan run have it too, as a hidden flag. Default fireflies. In the source: briar.extract._meetings.MEETINGS.

NameWhat it is
firefliesFireflies.ai. Reads FIREFLIES_ACME_API_KEY.

Chat tools

Where you use it: The hidden --chat flag on agent and plan run. Default slack. briar only reads chat; there is no scheduled chat extractor. In the source: briar.extract._chats.CHATS.

NameWhat it is
slackSlack, through a browser session. Reads SLACK_ACME_TOKEN and SLACK_ACME_COOKIE_D.

Agent, LLM and scaffold

The AI side of briar: which model company answers, which helper tools the agent always has, and the parts briar scaffold builds a config from.

LLM providers

Where you use it: --llm on plan build, plan next, plan run and chat. agent has no --llm: it always runs on Anthropic. plan build also takes an empty value, to build with no LLM. In the source: briar.agent._llms.LLMS.

NameWhat it is
anthropicClaude, by Anthropic. Reads CLAUDE_CODE_OAUTH_TOKEN or ANTHROPIC_API_KEY.
openaiOpenAI. Reads OPENAI_API_KEY. Needs the openai extra.
geminiGoogle Gemini. Reads GEMINI_API_KEY. Needs the gemini extra.
bedrockModels on AWS Bedrock (Claude by default), through your normal AWS login.

Built-in MCP servers

Where you use it: On in every agent run, including plan run and --dry-run, when pip install "briar-cli[mcp]" is done. Without the extra, briar skips them with one warning and the run goes on. Turn them off with --no-default-mcp or BRIAR_NO_DEFAULT_MCP=1. A runbook mcp: entry with the same name replaces one. In the source: briar.mcp.defaults.DEFAULT_MCP_SERVERS.

NameWhat it is
thinkStep by step thinking for hard tasks. Runs with npx. Without Node it is skipped with a warning.
timeThe date, the time and time zone math. Runs with uvx. Without uv it is skipped with a warning.

Scaffold templates

Where you use it: briar scaffold NAME. In the source: briar.iac.scaffold.TEMPLATES.

NameWhat it is
implementationFrom a ticket to a plan, a human approval, the code, and a PR. You can change each step.
pr-fixesReads PR review comments, pushes fixes and replies. No human approval step.

Agent roles

Where you use it: --archetype on both scaffold templates. In the source: briar.iac.scaffold.archetypes.ARCHETYPES.

NameWhat it is
engineerBuilds what a ticket asks for, commits, and opens a PR.
pr-ci-fixerFixes failing CI on open PRs, from the failure logs.
pr-conflict-resolverMerges the base branch into stale PRs, fixes the conflicts, and pushes.
pr-fixerAnswers open review comments with follow-up commits.
triagerSorts and labels tickets, and adds a short summary comment.

Workflow shapes

Where you use it: --shape on both scaffold templates. In the source: briar.iac.scaffold.shapes.WORKFLOW_SHAPES.

NameWhat it is
one-shotOne agent step. No approval, no branches.
plan-approve-actThe agent plans, a person approves, then it acts. On a reject, it comments.
triageOne agent that sorts and comments. It takes no write actions.

Triggers

Where you use it: --trigger-kind on both scaffold templates. In the source: briar.iac.scaffold.triggers.TRIGGER_TEMPLATES.

NameWhat it is
github_webhookStarts on GitHub webhook events, like issues.opened.
bitbucket_webhookStarts on Bitbucket webhook events, like issue:created.
schedule_cronStarts on a cron schedule. Hourly by default.
manualNo trigger. You start it by hand.

Sources

Where you use it: --source on both scaffold templates. In the source: briar.iac.scaffold.sources.SOURCE_TEMPLATES.

NameWhat it is
githubIssues from one repo. Tools to comment, open PRs and commit files.
bitbucketIssues and PRs from one Bitbucket Cloud repo. The same three tools.
jiraIssues from Jira projects. Tools to comment, change status and edit.
awsAWS state for the agent to read. No write tools.
sentryError groups from Sentry projects. Tools to comment, resolve, assign and ignore.

Plan boards

briar plan build reads cards from a board. You pass the board's link, and briar picks the reader that matches it.

Board readers

Where you use it: The board argument of briar plan build. In the source: briar.plan._boards.BOARD_READERS.

NameWhat it is
jiraA Jira board link, like https://acme.atlassian.net/jira/software/projects/ACME/boards/12, or the short form jira:ACME.
github-projectA GitHub Projects link, like https://github.com/orgs/acme-co/projects/3 (or /users/ for a personal project).

Storage and journal

Where briar keeps knowledge blobs and run records.

Knowledge stores

Where you use it: --store on most commands, store in .briar.toml, BRIAR_STORE, and knowledge.store in a runbook. In the source: briar.storage.KnowledgeStoreRegistry.

NameWhat it is
fileMarkdown files under ./knowledge. Nothing to set up.
postgresA Postgres database at BRIAR_DATABASE_URL, with history.

Journal stores

Where you use it: --store on briar journal, and BRIAR_JOURNAL_STORE. In the source: briar.journal.JOURNAL_STORE_NAMES.

NameWhat it is
fileOne JSON file per session under ./journal/sessions/.

Journal outputs

Where you use it: BRIAR_JOURNAL_SINKS, a comma list. An unknown name is skipped. In the source: briar.journal.sinks.JOURNAL_SINKS.

NameWhat it is
fileA readable markdown copy of each session in ./journal/published/.

Logins and keys

How briar gets keys and where it keeps them. Full details are in Auth, secrets and telemetry.

Login targets

Where you use it: The TARGET word of briar auth login, logout, refresh and status. In the source: briar.auth._acquirers.ACQUIRERS.

NameWhat it is
github-deviceGitHub, by typing a short code on github.com. Needs a GitHub OAuth app.
github-patGitHub, by pasting a personal access token.
bitbucket-app-passwordBitbucket Cloud, by pasting an app password.
aws-staticAWS, by pasting an IAM access key and secret.
aws-ssoAWS, through IAM Identity Center in the browser. Short-lived keys.
jira-tokenJira, by pasting an API token.
jira-sessionJira, by pasting a browser session cookie. For sites with no API tokens.
linear-api-keyLinear, by pasting a personal API key.
firefliesFireflies.ai, by pasting an API key.

Key stores

Where you use it: --cred-store on briar auth and briar secrets doctor, and BRIAR_DEFAULT_STORE. In the source: briar.credentials.STORES.

NameWhat it is
envfileA local file, ~/.config/briar/secrets.env by default. The default, and the only one loaded when commands run.
aws-secretsmanagerAWS Secrets Manager, as briar/NAME.
ssmAWS SSM Parameter Store, as /briar/NAME.
vaultHashiCorp Vault KV v2, at secret/briar/NAME. Needs the vault extra.

Startup loaders

Where you use it: --kind on briar secrets bootstrap. Each one that is set up also runs at the start of every command. In the source: briar.credentials._bootstraps.BOOTSTRAPS.

NameWhat it is
envfileLoads the env file into the env. Names you already set win.

Jira login kinds

Where you use it: JIRA_ACME_AUTH_KIND. When it is empty, briar picks session if a session cookie var is set, else token. In the source: briar.extract._trackers._jira_auth._JIRA_AUTHS.

NameWhat it is
tokenEmail and API token: JIRA_ACME_EMAIL and JIRA_ACME_TOKEN.
sessionBrowser cookies: JIRA_ACME_TENANT_SESSION_TOKEN or JIRA_ACME_SESSION_TOKEN.

Messages and alerts

Two ways briar posts outside. Messages are posts the agent makes while it works. Alerts are sent when a runbook task fails.

Message kinds

Where you use it: The kind: of each entry in a runbook company's messages: block. The agent posts with its send_message tool, only when it runs with --runbook. In the source: briar.messaging.WRITERS.

NameWhat it is
github-pr-commentA comment on a GitHub PR. Reads GITHUB_TOKEN.
bitbucket-pr-commentA comment on a Bitbucket PR or issue. Reads BITBUCKET_ACME_*.
jira-commentA comment on a Jira ticket. Reads JIRA_ACME_*.
jira-transitionMoves a Jira ticket to another status. Reads JIRA_ACME_*.
slack-channelA post in a Slack channel, by webhook. Reads SLACK_ACME_WEBHOOK_URL.
telegram-chatA post in a Telegram chat, by bot. Reads TELEGRAM_BOT_TOKEN and TELEGRAM_ACME_CHAT_ID.

Alert sinks

Where you use it: BRIAR_NOTIFY_SINKS, a comma list, like slack,email. An unknown name is logged and skipped. See Failure alerts. In the source: briar.notify.SINKS.

NameWhat it is
emailAn email over SMTP. Reads SMTP_*, EMAIL_FROM and EMAIL_ACME_TO.
pagerdutyA PagerDuty warning. Reads PAGERDUTY_ACME_ROUTING_KEY.
slackA Slack post by webhook. Reads SLACK_ACME_WEBHOOK_URL.
telegramA Telegram message by bot. Reads TELEGRAM_BOT_TOKEN and TELEGRAM_ACME_CHAT_ID.

Output formats

How commands print their results.

Formats

Where you use it: The global --format flag. It works anywhere on the line. Default table. In the source: briar.formatting.FORMATTERS.

NameWhat it is
tableA table for lists. A single record prints as JSON.
jsonJSON.
yamlYAML.
csvComma separated values, one row per line.
quietOnly the id of each row, one per line, for xargs.

See also