v0.5.0 — smart account selection + WorkflowState fix

[0.5.0] — 2026-06-24

Added

• Intelligent account auto-selection by team key. Accounts can now declare
  the team keys they own via teams = ["SWA", "ENG"] (config), the new
  clinear auth add --teams "SWA,ENG" flag, or clinear auth teams <name> "SWA,ENG".
  When a command targets a team — through --team SWA or an identifier like
  SWA-20 — clinear picks the owning account (and its token) automatically,
  with no --account flag or env-var juggling. Resolution order is now:
  --account › team-key ownership › workspace mapping › default › first.
• clinear auth teams <name> "<keys>" command to set/clear an account's team keys.
• teams shown in clinear auth accounts output (human + JSON).

Fixed

• WorkflowState enum no longer crashes on undocumented state types. Teams
  with a workflow state whose type is outside the six documented categories
  (e.g. "duplicate") previously broke clinear team states and
  clinear issue state with a Pydantic validation error. The type field is
  now a smart union (WorkflowStateType | str): known types become the enum,
  unknown types pass through as plain strings.

Changed

• Agent skill (SKILL.md + skill_content/) rewritten to match reality:
  corrected the multi-account/auth section (removed the dangerous "rewrite to
  [auth]" advice), fixed comment add/comment edit to positional body
  (no --body flag), documented --account and the new team auto-selection.

Tests

• New offline unit suite tests/test_account_resolution.py (20 tests) covering
  team_key_from_hint, resolve_account precedence, resolve_token, and the
  WorkflowState smart-union.

---

v0.4.1 — Memory Board for Agents

Added

• Project memory board (clinear memory) — persistent project-scoped memory
  for agents. Every agent invocation should start with clinear memory remind
  to load forced behavioral rules and recent community learnings.
  • memory remind — print a formatted digest (forced rules + community context)
  • memory list — list all entries with metadata
  • memory add --title "..." --body "..." — add a community entry
  • memory update <id> — edit an existing entry
  • memory remove <id> — delete an entry
  • memory heal --dry-run — remove stale community entries (older than
    heal_after_days, default 30)
  • Storage: .clinear/memory.yaml in the git repo root (project-scoped)
  • Seeded with 5 forced rules: verify auth, search before create, use JSON
    for piping, dry-run on ambiguous mutations, read memory board first
  • Community entries auto-heal after 30 days; forced entries never expire
• Cron-ready: clinear memory heal can be scheduled daily to keep the board
  clean and relevant.

Files Changed

• clinear/memory_board.py — storage layer (YAML read/write, heal logic)
• clinear/commands/memory.py — 6 subcommands
• clinear/cli.py — register memory_app
• clinear/skill_content/overview.md — add memory board rules

---

v0.4.0 — Multi-account credentials + Auto-update

Added

• Multi-account credential support — manage multiple Linear accounts and switch
  between them automatically per workspace or via explicit --account flag.
  • New config schema: [accounts.<name>] sections with token, token_env,
    and auto-populated org_name.
  • Workspace-aware defaults: git repository roots can be mapped to specific
    accounts via config.toml [workspaces] or automatically detected.
  • New clinear auth subcommands:
    • auth accounts — list all accounts with default/workspace/current markers
    • auth add <name> --token <token> — add a named account (optionally
      verifies token against Linear API to populate org_name)
    • auth switch <name> — set global default account
    • auth remove <name> — remove an account
    • auth workspace — show current git repo workspace and mapped account
  • New global flag: --account <name> / -a <name> — one-off override for
    any subcommand.
  • Backward compatibility: existing [auth] sections auto-migrate to
    accounts.default on first load. Single-account behavior is preserved
    as fallback.
• clinear update — self-update command that checks PyPI for the latest
  version and runs the appropriate upgrade command (pip install --upgrade or
  pipx upgrade). Supports --dry-run and --yes flags. JSON output mode
  supported for agent consumption.
• New dependency: tomli-w>=1.0 (required for writing config updates from
  auth add/switch/remove commands).

Changed

• config.py — replaced single AuthConfig with AccountConfig and
  AccountsConfig dict. Added resolve_account() for workspace-aware
  account selection and save_config() for persisting config changes.
• cli_state.py — added account_name field to CLIState for tracking
  the active account.
• cli.py — added --account global flag and update command registration.
• init.py — updated config template to use multi-account schema.

---

v0.3.1

v0.3.1

v0.3.0 — Agent skill + clinear-mcp server

Added

• Agent skill bundle at skills/clinear/ — a Swarm/Claude-style skill
  (SKILL.md frontmatter + references/*.md) that teaches AI agents both the
  mechanics of the CLI and the behavior of working through Linear (when to
  search before create, when to use --output json, how to chain commands,
  safety rules, anti-patterns). Install with bash skills/install.sh.
• clinear-mcp MCP server — optional Model Context Protocol server
  exposing:
  • 1 tool clinear_guide(topic) returning structured teaching content
    (overview / commands / workflows / filters / output-formats
    / examples).
  • 7 read-only resources: clinear://me, clinear://issue/{id},
    clinear://team/{key}, clinear://project/{id_or_slug},
    clinear://cycle/current/{team_key}, clinear://issues/mine,
    clinear://issues/team/{team_key}.
  • 6 prompt templates: triage, daily_standup, create_from_error,
    hand_off, cycle_review, issue_investigate.
• By design no mutation tools on the MCP server. Mutations happen via
  the clinear CLI in a shell — every tool/prompt response carries a
  reminder reinforcing this rule.
• clinear/skill_content/ — canonical markdown source bundled inside the
  wheel via [tool.hatch.build.targets.wheel.force-include]. Single source
  of truth for both the skill bundle and the MCP server.
• New optional dependency extra: pip install 'clinear[mcp]' pulls in
  mcp>=1.12.0. Core install stays untouched for users who don't need MCP.
• New console script clinear-mcp (entry point of the MCP server).

Changed

• Bumped version to 0.3.0.
• Updated User-Agent string emitted by LinearClient to clinear/0.3.0.

Notes

• The MCP code is lazy-imported. Users who install clinear without the
  [mcp] extra are not affected; the clinear-mcp script prints a friendly
  install hint and exits 2 if invoked without the dep.
• ALL MCP server logging is routed to stderr to keep stdio JSON-RPC clean.

---

v0.2.0 — Initial public release

Added

• clinear init — scaffold the config file at ~/.config/clinear/config.toml.
• clinear comment group: list, add, edit, delete.
• clinear label group: list, create, delete.
• Label resolution in issue create and issue update — accept label names
  (comma-separated) and resolve to UUIDs server-side per team.
• VERSION file at the repo root; __version__ now reads from it.
• CHANGELOG.md.

Changed

• cycle current now exits 0 with { "active_cycle": null } (JSON) or
  a friendly message (human) instead of crashing with NotFoundError when
  a team has no active cycle.
• issue search results now render a full table with priority, state,
  assignee, and identifier columns.
• Bumped version to 0.2.0.

Fixed

• 4 issues from the v0.1 smoke test:
  • Issue.labels / Issue.subscribers now flatten the GraphQL
    {nodes: [...]} connection wrapper automatically.
  • searchIssues no longer spreads the Issue fragment on
    IssueSearchResult (different GraphQL type).
  • YAML formatter no longer collapses nested object indentation.
  • Error exit codes propagate correctly through the typer entrypoint.

---