The 8genC MCP server is the secure channel between your AI agents and our operating playbook — an operating system of 78 Agent Skills across four levels, with orchestration, per-client memory, and scoped credentials, delivered over Model Context Protocol.
8genC pairs seasoned fractional leadership with agentic execution. The MCP server is how that execution gets standardized: it hands your AI agents the same ground truth and the same playbook our operators use — so the work comes back correct, compliant, and on-brand the first time.
Live, authoritative knowledge of the AINative platform — every service, API endpoint, SDK, and architectural constraint. Agents stop guessing at integrations and start citing real API paths.
A GitHub-backed library across four levels — service areas route to category leads, which orchestrate sub-agents and task agents (Business, Technology, Marketing, Sales). Loaded straight from source, exposed as MCP tools + prompts, and resolvable into parallel/dependent runs.
Skills are authored in GitHub and mirrored into ZeroDB for semantic search. Edit the playbook once; every connected agent — across every client engagement — pulls the same current version.
For 8genC operators running client engagements. Spin up agents that already know our delivery standards — generate PRDs, suggest stacks, and build agentic systems against the exact templates and validation rubrics the firm runs on.
For client teams operating inside an 8genC engagement. Connect your own agents to a scoped channel that carries your project's context — and execute against the same playbook your 8genC operators use, with full visibility.
Register at the access portal and verify your email. An admin approves your account and assigns your track — consultant or client.
Open /access →Sign in via OAuth at connect time, or create a personal access token in the portal and add it as a Bearer header in your MCP client.
Auth methods →Drop the endpoint into Claude, Cursor, or Windsurf. The 16 tools and 78 Agent Skills appear instantly — ready to execute.
Full manual →
prd_list_servicesList all 22 AINative products with APIs, SDKs & pricing.
prd_get_api_catalogDetailed API info for a specific service.
prd_suggest_stackRecommend services for a requirement, with justification.
skill_listList Agent Skills pulled live from GitHub.
skill_getFetch a skill's full SKILL.md body.
skill_get_referencePull a single reference file on demand.
skill_searchSemantic search for the right skill for a task.
skill_syncMirror GitHub skills into ZeroDB for offline use.
Two real-shaped runs — one internal, one client-facing — showing how the tools chain into a finished deliverable.
Every request to POST /mcp must carry a valid credential for an approved account. No approval, no channel.
The server is its own authorization server. Compatible clients open a browser, you sign in and consent, the client receives a scoped token.
Mint 8genc_pat_… tokens in the portal and pass them as a Bearer header. Revoke any token, any time.
New accounts stay pending until an admin approves and assigns a track. Block, unblock, and reset from the portal.
Email verification on every account, optional Google / GitHub social sign-in, JWT-backed sessions.
For the hosted endpoint at mcp.8genc.com/mcp — yes. Every request is gated on an approved account. Register at the access portal, verify your email, and an admin approves you.
Run locally over stdio and platform discovery plus the core skill tools work with zero credentials — only semantic search and sync need a ZeroDB account.
Any client that speaks Streamable HTTP MCP — Claude, Cursor, and Windsurf are the common ones. Clients that support MCP OAuth get browser sign-in at connect; everything else uses a personal access token as a Bearer header.
The source of truth is the GitHub repo the8genc/ai-8gent-skills, laid out as skills/<slug>/SKILL.md plus reference files. ZeroDB is a cache and semantic-search layer — author skills in GitHub, then skill_sync to refresh the mirror.
PRD authoring moved out of the server and into the 8gentic-prd Agent Skill. The skill carries the full workflow, templates, and validation rubric — and orchestrates the server's platform tools and your ZeroDB memory tools rather than re-implementing them. Load it with skill_get 8gentic-prd.
Both personas connect to the same channel and the same playbook — the distinction is scope and approval. Admins assign a track on approval; client accounts are scoped to their engagement, consultant accounts carry the firm's internal context. Same standards, different blast radius.