@leadbay/mcp 0.21.3 → 0.23.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +18 -0
- package/dist/bin.js +468 -4
- package/dist/http-server.js +442 -3
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,23 @@
|
|
|
1
1
|
# Changelog — @leadbay/mcp
|
|
2
2
|
|
|
3
|
+
## 0.23.0 — 2026-06-21
|
|
4
|
+
|
|
5
|
+
Guided campaign builder.
|
|
6
|
+
|
|
7
|
+
- **`leadbay_build_campaign`** (new prompt + auto-generated Claude skill) — one guided on-ramp from "build me a campaign" to a ready-to-work campaign: discover on the active lens → qualify → pick an ICP pool → enrich the BUYER PERSONA of the user's product (a sales tool's buyer is the revenue org — VP/Head/Dir Sales, BD, CRO, CMO, growth — not whoever is most senior) → create the campaign → render the `leadbay_campaign_call_sheet` view → hand off to `leadbay_work_campaign`.
|
|
8
|
+
- **Coverage guarantee** — drops/swaps any lead with no enrichable buyer-persona contact for the highest-score in-ICP lead that has one (never trading ICP for coverage), and a ⚠ flag for suspect emails (domain ≠ company, or a contact on >1 lead). Polls enrichment to completion before rendering.
|
|
9
|
+
- Pure orchestration of existing composites — no new tool. Eval workflow #34 added; tuned over 4 live eval cycles (campaign quality 3/7 → 11/11 leads with a real revenue buyer reachable).
|
|
10
|
+
|
|
11
|
+
## 0.22.0 — 2026-06-20
|
|
12
|
+
|
|
13
|
+
Headless artifact SDK + two always-on read tools, so the user's Claude (cowork) can build interactive HTML artifacts that call Leadbay.
|
|
14
|
+
|
|
15
|
+
- **New package `@leadbay/components`** — vanilla view-models (`lb.field` / `action` / `resource` / `list`, plus domain helpers `outreach`, `note`, `like`/`dislike`, `leadHistory`, `leadProfile`, `callList`, `enrichment`, `teamActivity`) that own a control's data lifecycle: populate-from-API, value/loading/error, validation, polling, request sequencing, a 30s call timeout, and the `report_outreach` verification + `_triggered_by` footguns. The agent owns 100% of markup; the library renders nothing. The build emits a minified runtime into core; a `components:check` drift guard fails CI if the committed runtime goes stale.
|
|
16
|
+
- **`leadbay_artifact_kit`** (always-on read) — returns the runtime string + a usage guide the agent reads to assemble an artifact. Lives in `tools/` (granular-shaped) so a kit fetch carries no `_triggered_by` mandate.
|
|
17
|
+
- **`leadbay_team_activity`** (always-on read) — per-rep activity leaderboard + activity trend for a look-back window, wrapping `/kpi/users` + `/kpi/trends` (the web Dashboard-Manager data). Admins get the whole org; non-admins are scoped to themselves by the backend.
|
|
18
|
+
|
|
19
|
+
Validated by agent-dogfood tests: an independent agent builds a working cold-call sheet + a manager dashboard from the usage guide alone, and jsdom asserts the real tool calls fire with the right args. (Core 0.8.4, components 0.3.1.)
|
|
20
|
+
|
|
3
21
|
## 0.21.3 — 2026-06-19
|
|
4
22
|
|
|
5
23
|
Kills the 401 startup "reconnect Leadbay" hallucination — the assistant told users to re-authenticate on a connection that actually worked (product#3761). Fixed at every layer that produced or surfaced the spurious 401:
|