Jakob WennbergandClaude Opus 4.8
6be6510d73
* fix(entitlements): gate paid AI document OCR server-side (free-tier leak) Free/manual-tier companies could trigger paid Bedrock OCR (extractInvoiceFields) with no `ai` capability check, on every transport: - invoice-inbox HTTP paths — /upload + email /inbound (shared uploadAndExtract), /items/:id/attach-document, /items/:id/retry-extraction (4 call sites, zero capability refs); - the gnubok_upload_document MCP tool — absent from MCP_TOOL_CAPABILITY_MAP, so a free-tier API key (incl. the claude.ai connector's minted gnubok_sk_ key) got unlimited AI extraction. This disproved the keys.ts "no MCP tool invokes AI" comment. Fix (money-blocker for the free/paid tier cutover): - Gate the 3 inbox call sites on hasCapability(CAPABILITY.ai). Upload + attach degrade gracefully (document still stored; extraction skipped with reason `no_ai_entitlement`, highest priority in the existing skipReason chain). Retry is an explicit "run AI now" action, so it hard-blocks with 403 capabilityBlockedResponse. - Register gnubok_upload_document -> CAPABILITY.ai in MCP_TOOL_CAPABILITY_MAP; the dispatcher already enforces the map. Correct the stale keys.ts comment and the misleading "deterministic field extraction" tool description + manifest copy (the extension migrated regex -> AI OCR). Tests: no-AI upload/attach skip + retry 403 (sandbox-skip-extraction), retry 403 (retry-extraction), and the MCP map contract + refined dispatch<->commit parity (capability-maps: upload_document is dispatch-only, no commit counterpart). Self-hosted stays all-on (hasCapability short-circuits). No migration. Follow-up (not in scope): capability-blind DashboardNav (free/paid rails identical) and /chat gated on isVerified not `ai` — see dev_docs/nav_ia_redesign.md Part 4. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(mcp): assert gnubok_upload_document is ai-gated at dispatch The gnubok_upload_document handler runs extractInvoiceFields (Bedrock OCR) inline rather than through the entitlement-gated uploadAndExtract, so the central MCP_TOOL_CAPABILITY_MAP dispatch check is the only paywall on that transport. Lock it with a test (flagged by PR review as an untested money path) so a free-tier connector key can never reach paid OCR. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <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