Files
accounted/extensions/general/mcp-server
MattssonandClaude Fable 5 1fa34aa7ca feat(skatteverket): repair notification recipients + make the agent the SKV notification surface (#1887)
* feat(skatteverket): repair notification recipients + make the agent the SKV notification surface

The company_members -> profiles!inner(email) PostgREST embed has no FK to
traverse (company_members.user_id references auth.users), so it 400'd and
silently killed all four notification emails since they shipped. Recipient
lookup is now a shared two-step helper (lib/notifications/member-email):
kvittens confirmations, skattekonto drift alerts (tax-contact routing
preserved via the plural variant) and backup alerts deliver again. The
connection-expired email is deleted instead of fixed: with SKV's 65-minute
personal sessions it was one mail per connect (see DECISIONS.md); the event
and needs_reconsent flagging stay.

For MCP-first users the agent is the notification surface, so:
- SKATTEVERKET_NOT_CONNECTED copy is now agent-directive: session expiry is
  normal (~1h by SKV design), only a person can reconnect with BankID, do
  not retry until they confirm. Inline strings (declaration-status, read
  routes, v1 pitfalls, accounted-api skill) aligned.
- gnubok_get_agent_briefing gains an optional skatteverket_connection block
  (status/source/connected_at + directive message on needs_reconsent),
  emitted only when a connection or verified system grant exists, so agents
  warn the user at session start instead of failing mid-task. Payload bench
  ceiling bumped 59.95K -> 60.15K for the outputSchema contract.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(skatteverket): drift email resolves recipients via service client; review fixes

The skeptic pass refuted the drift-email repair: skattekonto.drift_detected
is emitted only by the nightly cron, and the extension registry builds each
event handler a fresh ctx from the anonymous cookie client (or none at all
on cookieless requests), so RLS returned zero company_members rows and the
two-step lookup still resolved no recipient. The handler now builds its own
service-role client, the same documented pattern as the retired
connection-expired handler; drift tests exercise the handler without ctx,
matching the cron reality.

CodeRabbit findings: resolveMemberEmails pages both queries through
fetchAllRows with stable ordering (PostgREST caps unpaged reads at 1000
rows); the v1 vat-declarations pitfall and regenerated accounted-api docs
now name both auth paths (member BankID connection or verified ombud
grant); the briefing's system-before-user priority carries a cross-reference
to resolveReadAuth explaining why it is not reused. member-email.ts JSDoc
states the service-role-client requirement (profiles RLS is own-row-only).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2026-08-25 12:09:20 +02:00
..
2026-07-24 15:03:50 +02:00
2026-07-24 15:03:50 +02:00
2026-07-24 15:03:50 +02:00

Accounted MCP server

JSON-RPC 2.0 server exposing the Accounted bookkeeping engine to MCP clients (Claude Desktop, Claude Code, etc.). Endpoint: /api/extensions/ext/mcp-server/mcp. Add ?tool_namespace=accounted for the Accounted tool names. Requests without it retain the legacy Gnubok namespace. OAuth and stdio bridges live alongside the API surface: see app/api/mcp-oauth/, packages/accounted-mcp/, and the compatibility package in packages/gnubok-mcp/.

Tool authoring contract

Enforced by tests in __tests__/: these are not style preferences, they're guard rails.

  1. additionalProperties: false on every inputSchema. Guarded by strict-schemas.test.ts. Forces clear rejections on hallucinated fields instead of silent ignores.
  2. Descriptions ≤ 280 chars. Guarded by output-schema.test.ts. No Args: / Returns: / Examples: prose: those belong in JSON Schema. Use agent-native hints ("Use to…", "Call X first", "HIGH risk").
  3. Staged-operation envelope for write tools: outputSchema: STAGED_OPERATION_SCHEMA (server.ts). Fields: staged, risk_level, actor, message, preview, period_status?, next?. The staged: true boolean is the explicit completion signal; agents must not infer completion from prose. Do NOT introduce a parallel { success, shouldContinue, output } envelope.
  4. period_status threading: any tool that ties to a fiscal-period-bound date (categorize, mark paid, create voucher, correct/reverse entry, approve supplier invoice) passes dateForPeriodCheck to stagePendingOperation. Response then includes period_status: { period_id, status: open|locked|closed, lock_date } so widgets and agents disable writes without round-trips.
  5. Scope mapping: every new tool needs an entry in lib/auth/api-keys.ts TOOL_SCOPE_MAP. Missing entries default to deny.
  6. Tests for new write tools: add staging-gate coverage to __tests__/voucher-tools.test.ts (or a sibling) plus executor coverage to lib/pending-operations/__tests__/voucher-executors.test.ts if the tool stages a new operation_type.

Determinism / cache stability

Tool definitions (name, description, inputSchema, outputSchema, annotations) are declared as static object literals at module load: no timestamps, no UUIDs, no Date/Math.random in the definition layer. This makes the tools/list JSON payload byte-stable across requests, which lets agent-side prompt caches stay warm. Do not introduce per-request non-determinism into the definitions block. Anything time-bound or random belongs inside execute().

For internal Anthropic API usage (today only extensions/general/invoice-inbox/lib/extract-invoice-fields.ts): annotate stable prefixes with cache_control: { type: 'ephemeral' } and log usage.cache_read_input_tokens for hit-ratio observability. The 1h TTL from the agent-native API plan (item 10) requires the direct Anthropic API; Accounted's Bedrock path defaults to a shorter TTL.

Payload-size watchdog

payload-size.bench.test.ts enforces a tools/list JSON payload ceiling. If the test fires, the right answer is rarely "raise the ceiling". Instead, trim descriptions or set specialized wide tools to catalogVisibility: 'search'. Those tools remain discoverable with full schemas through gnubok_search_tools and callable through tools/call without bloating the default catalog.

Where things live

  • server.ts: the tools array + JSON-RPC dispatcher
  • tool-result.ts: withNext(), toToolError() response helpers
  • resources/: read-only Accounted:// URIs (active company, period, recent activity, capabilities, attention items, voucher gaps, chart of accounts, VAT treatments)
  • widgets/: inline HTML widgets (receipt-matcher, vat-review)
  • prompts/: slash-command-style prompts
  • skills/: domain-knowledge skill bodies served via gnubok_load_skill
  • __tests__/: strictness guards + per-tool coverage