Problem
Creating multiple entries via MCP requires sequential `kb_create` calls, each triggering an individual index sync and embedding generation. For batch operations (e.g., importing a set of actors or events), this is slow and produces excessive MCP round-trips.
Solution (Implemented)
Added `kb_bulk_create` write-tier MCP tool with best-effort per-entry semantics:
```
kb_bulk_create:
kb_name: str
entries: array of {entry_type, title, body, date, importance, tags, metadata}
```
Behavior
Max 50 entries per call
Best-effort: each entry succeeds or fails independently
Returns per-entry results: `{created: true, entry_id}` or `{created: false, error}`
Summary counts: `{total, created, failed, results}`
KB registration happens once (not per-entry)
Embedding is batched after all entries are indexedFiles Modified
`pyrite/services/kb_service.py` — `bulk_create_entries()` method
`pyrite/server/mcp_server.py` — `kb_bulk_create` tool schema + `_kb_bulk_create` handler
`tests/test_mcp_server.py` — 5 tests (happy path, partial failure, empty, read-only, over-limit)