Jakob WennbergandClaude Fable 5
678f2ccffd
* feat(mcp): counterparty-tied category suggestions + no_signal (P2-1)
suggest_categories padded every transaction with a company-wide
category-frequency fallback at <=0.5 confidence — an identical four-way
spread on 20+/24 items that agents correctly reported as pure noise
(agent.feedback). Real signal came from memory atoms and query_journal.
- History is now counterparty-keyed: buildMerchantHistory groups past
categorized transactions by normalized merchant; the engine only
surfaces history for THIS transaction's merchant, with provenance
('Bokförd N gånger tidigare för denna motpart') and occurrence-scaled
confidence (0.56 at 1x, capped 0.85). No global padding — an empty
list is the honest answer.
- The MCP tool returns no_signal_transaction_ids for transactions where
NO source matched, steering agents to investigate (query_journal)
instead of pattern-matching on unrelated rows.
- Both callers (REST suggest-categories route + MCP tool) share the new
helpers, so web UI and agents improve together.
Part of dev_docs/mcp_optimization_plan.md (P2-1).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(skills): dangling-reference validation in skills:check + fix 10 dangling links (P2-2)
skills:generate/check now fail when an atom SKILL.md links a
references/*.md that does not exist on disk — a dangling pointer ships
a 404 to every agent that follows it (the weekly-booking-check
incident, agent.feedback).
The validator immediately caught 10 live dangling links in 4 atoms,
three distinct flavors:
- filename typo: swedish-asset-accounting/references/depreciaton.md
renamed to depreciation.md (the link was right, the file misspelled)
- link mismatch: swedish-e-invoicing linked market-providers-pricing.md;
the file is market-provider-pricing.md (link fixed)
- unauthored plans: single-shareholder-ab-fmb TODOs and reklambyra's
'planerad utbyggnad' section used resolvable references/ paths for
files that were never written — rephrased as plans without paths
Seed migration regenerated (4 atoms bumped, renamed reference child).
Part of dev_docs/mcp_optimization_plan.md (P2-2).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs(events): align agent-feedback review cadence copy (P2-4)
gnubok_feedback replies 'we aggregate signal weekly'; the event-log
handler comment said quarterly. One of them was lying — weekly wins
(the mcp_optimization_plan triage is the living example).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
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. OAuth and stdio bridge live alongside the API surface — see app/api/mcp-oauth/ and packages/gnubok-mcp/.
Tool authoring contract
Enforced by tests in __tests__/ — these are not style preferences, they're guard rails.
additionalProperties: falseon everyinputSchema. Guarded bystrict-schemas.test.ts. Forces clear rejections on hallucinated fields instead of silent ignores.- Descriptions ≤ 280 chars. Guarded by
output-schema.test.ts. NoArgs:/Returns:/Examples:prose — those belong in JSON Schema. Use agent-native hints ("Use to…", "Call X first", "HIGH risk"). - Staged-operation envelope for write tools —
outputSchema: STAGED_OPERATION_SCHEMA(server.ts). Fields:staged, risk_level, actor, message, preview, period_status?, next?. Thestaged: trueboolean is the explicit completion signal; agents must not infer completion from prose. Do NOT introduce a parallel{ success, shouldContinue, output }envelope. period_statusthreading — any tool that ties to a fiscal-period-bound date (categorize, mark paid, create voucher, correct/reverse entry, approve supplier invoice) passesdateForPeriodChecktostagePendingOperation. Response then includesperiod_status: { period_id, status: open|locked|closed, lock_date }so widgets and agents disable writes without round-trips.- Scope mapping — every new tool needs an entry in
lib/auth/api-keys.tsTOOL_SCOPE_MAP. Missing entries default to deny. - Tests for new write tools — add staging-gate coverage to
__tests__/voucher-tools.test.ts(or a sibling) plus executor coverage tolib/pending-operations/__tests__/voucher-executors.test.tsif the tool stages a newoperation_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 (currently 36,000 tokens — bumped from 32,000 when top-level Tool.title landed on all tools for Claude Connectors Directory readiness). If the test fires, the right answer is rarely "raise the ceiling" — instead, trim descriptions or leverage gnubok_search_tools (already deployed; tool definitions can defer to it for discovery rather than enumerating in tools/list).
Where things live
server.ts— the tools array + JSON-RPC dispatchertool-result.ts—withNext(),toToolError()response helpersresources/— read-onlyAccounted://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 promptsskills/— domain-knowledge skill bodies served viagnubok_load_skill__tests__/— strictness guards + per-tool coverage