Use cases & roadmap
GitHub & engineering assistants
Ground agents on internal docs and call GitHub MCP tools for commits, files, and repo context—without pasting tokens into chat.
GitHub & engineering assistants
Use this pattern when engineering, DevOps, or release teams want answers that combine your written standards (in Knowledge) with live repository context (via GitHub MCP).
What ships today
| Capability | Where |
|---|---|
| GitHub MCP | Dashboard → Connectors → GitHub (see MCP — GitHub) |
| Knowledge search | Upload runbooks, ADRs, and onboarding PDFs; agent searches before answering |
| Playground approvals | Sensitive tool runs can pause for review in Agent chat |
| Workflows | Automate repo checks on webhook (e.g. MCP tool step after Input) |
Common GitHub MCP tools after you connect:
get_commit_history— recent commits forowner/reporead_repository_files— file contents at a refanalyze_repository— high-level structure signals
Example agent: Release helper
- Agents → Create — name:
Release Helper. - Instructions — cite repo facts from tools; never invent merge permissions; escalate if unsure.
- Knowledge — upload
RELEASE.md,CONTRIBUTING.md, and security checklists to aengineering/folder. - Tools — enable Knowledge search; enable GitHub connector (MCP).
- Guardrails — block requests for secrets, production deploys without approval.
Prompts to test in playground
- “What changed in
owner/repoonmainin the last 7 days?” - “Summarize our release checklist and tell me what’s missing for v2.4.”
- “Read
README.mdfromowner/repoatmainand explain the setup steps.”
Expected behavior:
- Answers ground policy text in Knowledge when relevant.
- Repo facts come from GitHub tools, not hallucination.
- Write actions (if any) stay behind approval or escalation.
Optional: workflow for release webhooks
For CI or ticketing systems that POST JSON:
- Input —
repo,branch,ticketId. - MCP tool —
get_commit_historywith args from inputs. - LLM — short release note draft.
- Output — return JSON for your notifier.
Trigger via workflow webhook or dashboard API.
SDK setup (TypeScript)
import { WisebotAI, Tools } from "@wisebotai/sdk";
const client = new WisebotAI({
apiKey: process.env.WISEBOTAI_API_KEY ?? process.env.WISE_KEY!,
baseUrl: process.env.WISEBOTAI_BASE_URL,
});
await client.agents.create({
name: "Release Helper",
model: "gpt-4o-mini",
tools: [
Tools.KnowledgeBase({ label: "RELEASE.md" }),
Tools.Connector({ provider: "github" }),
],
});Connect GitHub in the dashboard before expecting live tool calls.