Reference

Usage flows.

Recipes that chain several briar commands into a real outcome. Most flows deliberately combine more than one feature — a typical shape is secrets (gate) → extract (mine) → agent/plan (act) → context (inspect) → journal/dashboard (audit). Each flow names the Features combined so you can see what's working together. Flow 7 is the capstone that strings the whole tool into one chain.

Placeholders

<COMPANY> (e.g. acme) · <OWNER> (GitHub org / Bitbucket workspace) · <REPO> · <PROJECT> (tracker project key). Per-flag detail lives in the CLI reference.

1. Onboard a brand-new company from zero

Get credentials in place, prove coverage, take the first cold snapshot.

# 1. auth — acquire per-provider credentials
$ briar auth login github-pat --company <COMPANY>
$ briar auth login jira-token --company <COMPANY>
$ briar auth login aws-static --company <COMPANY>
$ briar auth login fireflies  --company <COMPANY>   # paste your Fireflies API key
 
# 2. secrets — prove every (company, extractor) pair is covered
$ briar secrets doctor --examples examples/
 
# 3. extract — first cold snapshot
$ briar extract --company <COMPANY> \
    --pr-repo <OWNER>/<REPO> --ticket-project <PROJECT>

Features combined: auth · secrets · extract

2. Extract AWS + Fireflies + PRs, then fix a PR

The headline flow: mine infra, meeting decisions, and PR history into the knowledge blob, then run the PR-fixer agent with the relevant Fireflies transcript spliced in just-in-time.

# 1. secrets — gate before spending anything
$ briar secrets doctor --examples examples/
 
# 2. extract — three sources into knowledge:<COMPANY>
$ briar extract --company <COMPANY> \
    --include aws-infra --include meeting-digest --include pr-archaeology \
    --pr-repo <OWNER>/<REPO> \
    --aws-extract-region us-east-1 --meeting-since-days 14
 
# 3. context — eyeball what landed before the agent reads it
$ briar context get knowledge:<COMPANY> | head -40
 
# 4. agent — address PR #128's open review comments (add --dry-run to preview)
$ briar agent prfix --company <COMPANY> --owner <OWNER> --repo <REPO> \
    --pr 128 --branch fix/login-retry \
    --meeting fireflies --meeting-query "login retry" --meeting-top-k 3 \
    --runbook examples/<COMPANY>.yaml
 
# 5. journal — audit the agent's decision trail
$ briar journal list --command-prefix agent.
$ briar journal show <SESSION_ID>

Features combined: secrets · extract (3 extractors) · context · agent prfix (+ JIT meeting context) · journal

Preview for free first

Add --dry-run to step 4 to render the exact system prompt + tool list (including the spliced ## Meeting context section) without an LLM call.

3. Implement a Jira ticket end-to-end

Clone, branch, code, test, open a draft PR — one ticket, autonomously.

# 1. secrets — confirm tracker + repo creds
$ briar secrets doctor --examples examples/
 
# 2. extract — fresh conventions + ticket context for the engineer agent
$ briar extract --company <COMPANY> \
    --include codebase-conventions --include active-tickets \
    --pr-repo <OWNER>/<REPO> --ticket-project <PROJECT>
 
# 3. agent — implement one ticket (ticket-context fetched JIT for the key)
$ briar agent implement --company <COMPANY> --owner <OWNER> --repo <REPO> \
    --ticket-project <PROJECT> --ticket-key <PROJECT>-412 --tracker jira \
    --runbook examples/<COMPANY>.yaml
 
# 4. journal — read back exactly what the agent did and why
$ briar journal show <SESSION_ID>

Features combined: secrets · extract · agent implement (+ JIT ticket context) · journal

For many tickets at once, graduate to the plan loop in Flow 4.

4. Build and run an implementation plan

Turn a tracker board into an ordered plan, then let the selector→implement→writeback loop ship it card by card while you watch the plan-knowledge blob learn.

# 0. extract — refresh the knowledge the synthesiser splices in
$ briar extract --company <COMPANY> --include codebase-conventions \
    --include active-tickets --pr-repo <OWNER>/<REPO> --ticket-project <PROJECT>
 
# 1. plan build — board → ordered plan, knowledge spliced in
$ briar plan build "https://github.com/orgs/<OWNER>/projects/7" \
    --company <COMPANY> --name q3-auth --with-knowledge \
    --llm anthropic --store postgres
 
# 2. plan — inspect before spending money
$ briar plan status q3-auth --company <COMPANY> --store postgres
$ briar plan next   q3-auth --company <COMPANY> --store postgres
 
# 3. plan run — smoke ONE card end-to-end
$ briar plan run q3-auth --owner <OWNER> --repo <REPO> \
    --tracker jira --tracker-project <PROJECT> \
    --llm anthropic --company <COMPANY> --store postgres --limit 1
 
# 4. context — watch the plan-scoped knowledge blob the loop updates
$ briar context --store postgres get knowledge:<COMPANY>.q3-auth | tail -20
 
# 5. plan run — go wide; keep going past a failing card
$ briar plan run q3-auth --owner <OWNER> --repo <REPO> \
    --tracker jira --tracker-project <PROJECT> \
    --llm anthropic --company <COMPANY> --store postgres --continue-on-failure
 
# 6. dashboard + journal — monitor + audit
$ briar dashboard --examples examples/ --once
$ briar journal list --command-prefix plan.

Features combined: extract · plan build/status/next/run · context (plan-knowledge blob) · dashboard · journal

5. Account-wide AWS inventory, on demand

Enumerate every tagged resource and keep a queryable, drift-tracked JSON companion — the prompt blob stays small (counts only).

# Option A — ad-hoc, full inventory dumped to a JSON sidecar
$ briar extract --company <COMPANY> --include aws-infra \
    --aws-extract-service tagging-inventory \
    --aws-extract-region us-east-1 --out-json /tmp/<COMPANY>-aws.json
 
# Option B — scheduled, persisting the companion automatically.
#   In the runbook YAML, on the company's knowledge binding:
#     knowledge: { store: postgres, name: knowledge:<COMPANY>,
#                  config: { inventory: "true" } }
$ briar runbook extract examples/<COMPANY>.yaml
 
# Inspect the companion and watch it drift over time
$ briar context --store postgres list --prefix inventory:
$ briar context --store postgres get inventory:<COMPANY> \
    | jq '.sections[].data.resources | length'

Features combined: extract / runbook · context (inventory companion)

6. Cost-safe agent rollout

Three escalating gates so you never discover a misconfiguration at scale.

# 1. FREE — render the prompt + tools, skip the LLM
$ briar agent implement --company <COMPANY> --owner <OWNER> --repo <REPO> \
    --ticket-project <PROJECT> --ticket-key <PROJECT>-77 --tracker jira --dry-run
 
# 2. ONE paid card through the plan loop
$ briar plan run q3-auth --owner <OWNER> --repo <REPO> \
    --tracker jira --tracker-project <PROJECT> \
    --llm anthropic --company <COMPANY> --limit 1
 
# 3. GO WIDE
$ briar plan run q3-auth --owner <OWNER> --repo <REPO> \
    --tracker jira --tracker-project <PROJECT> \
    --llm anthropic --company <COMPANY> --continue-on-failure

Features combined: agent implement (dry-run) · plan run (limited → wide)

7. Full lifecycle in one sitting (capstone)

Every major feature in one chain — from no credentials to merged AI-authored PRs with an audit trail.

# 1. auth + secrets — credentials in, coverage proven
$ briar auth login github-pat --company <COMPANY>
$ briar auth login jira-token --company <COMPANY>
$ briar auth login aws-static --company <COMPANY>
$ briar auth login fireflies  --company <COMPANY>
$ briar secrets doctor --examples examples/
 
# 2. runbook — scheduled extraction keeps knowledge fresh
#    (knowledge.config.inventory: "true" also persists the AWS companion)
$ briar runbook sweep examples/          # one-shot now
# briar runbook serve examples/        # ...or the daemon
 
# 3. context — confirm the knowledge + inventory blobs exist
$ briar context --store postgres list --prefix knowledge:
$ briar context --store postgres list --prefix inventory:
 
# 4. plan — board → ordered plan, knowledge spliced in
$ briar plan build "https://github.com/orgs/<OWNER>/projects/7" \
    --company <COMPANY> --name q3-auth --with-knowledge \
    --llm anthropic --store postgres
 
# 5. plan run — ship card by card
$ briar plan run q3-auth --owner <OWNER> --repo <REPO> \
    --tracker jira --tracker-project <PROJECT> \
    --llm anthropic --company <COMPANY> --store postgres --continue-on-failure
 
# 6. agent prfix — address review comments on a shipped PR
$ briar agent prfix --company <COMPANY> --owner <OWNER> --repo <REPO> \
    --pr 131 --branch q3-auth/card-3 \
    --meeting fireflies --meeting-query "auth review" \
    --runbook examples/<COMPANY>.yaml
 
# 7. dashboard + journal — monitor the estate, audit every decision
$ briar dashboard --examples examples/ --once
$ briar journal list --command-prefix plan.
$ briar journal list --command-prefix agent.

Features combined: auth · secrets · runbook · extract · context · plan · agent prfix · dashboard · journal — the whole tool in nine commands.

Where to go next

Recipes — flag-level worked examples for every command surface.
briar extract — the source of truth for extractors and flags.
Runbook YAML schema — every field, including knowledge.config.inventory.