Context Sites — agent guide

1. What it is

Context Sites (the Context Blog MCP) is the publishing half of Context: it drafts and publishes blog posts, site pages and Instagram carousels for a tenant, renders previews, lints content, records checker verdicts, and reports traffic and AI visibility. It never tracks work or approves anything — every piece carries a context_issue_id, and the approval that unlocks publish lives in Context.

2. Connect

• Endpoint: https://sites.onecontext.me/api/mcp (contract blog-1.3.0). The old blog.onecontext.me/api/mcp answers 410 moved.
• Connect both servers: Context first (https://mcp.onecontext.me/mcp — see the Context guide), then this one.
• Claude Code: claude mcp add --transport http context-blog https://sites.onecontext.me/api/mcp
• claude.ai: Settings → Connectors → Add custom connector → name Context Blog, URL https://sites.onecontext.me/api/mcp
• ChatGPT: Settings → Connectors → Add → URL https://sites.onecontext.me/api/mcp
• Cursor: add an HTTP MCP server with URL https://sites.onecontext.me/api/mcp
• Auth: Google sign-in for interactive hosts (no key). Headless routines use a bearer by name (CONTEXT_BLOG_MCP_TOKEN in the host's secret store, expanded into the Authorization header by the MCP config). Tenant credentials (Postiz, analytics, search console) are referred to by key name only.

3. First call

1. usage_guide — what the server is, tiers (artifact | review | hosted), rules. Then get_capabilities to confirm the tools this contract serves; if a tool a skill names is absent, say "not available yet on this server" and attach the payload you would have sent to the Context Issue instead.
2. setup {workflow: "blog" | "site" | "instagram"} — resolves the workflow skill (blog-agent / site-builder / instagram-drafter) plus rules-blog and dependencies, and returns mode: "files" (paths to write per harness + an AGENTS.md block) or mode: "chat" (skill links + inline protocol), with a nextStep. Idempotent. Do the nextStep.
3. list_skill_catalog / get_skill {key} read the same catalog as https://agents.onecontext.me/skills.

4. Vocabulary

• Tenant — one site (for example getmeetly.ai/blog); tenant_slug on every content call. Tier artifact (drafts only), review, hosted (served on a domain).
• Brand / audiences / hubs / topics — voice, byline, design tokens; audience segments; topic hubs (every article belongs to one, hub_slug); topics are the research backlog with target_query and fanout_queries.
• Article / page / instagram_post — the content rows. Status moves one step forward only: draft → in_review → approved → published → retired. Articles carry slug, locale, translation_group (= the EN master's slug), sections, faq, seo.
• context_issue_id — the Context Issue that tracks the piece. Required on article_upsert, page_upsert, instagram_post_upsert and publish.
• Checker verdict — pass | bounce | escalate, recorded with check_record; the independent blog-checker writes it, not the maker.
• Assets — images uploaded per tenant (asset_upload → PUT → asset_complete) and referenced by id (Instagram slides) or public URL.

5. The loop

1. Set up — setup {workflow}; for a new tenant tenant_create, brand_upsert, audiences_upsert, hubs_upsert, topics_upsert; the workflow skill charts the tenant Epic in Context (one parent Issue per channel, every piece a child).
2. Draft — in Context, claim the publish Issue (update_issues … in_progress). Research with cited sources. content_lint first (read-only; same body as the upsert), fix every finding, then article_upsert {tenant_slug, slug, locale, translation_group, title, description, sections, seo, hub_slug, context_issue_id} — it lints again and refuses on failure. Two bounces, then escalate.
3. Attach for review — on the Context Issue: the markdown deliverable (what the owner comments on) and ONE final preview: preview_render {tenant_slug, kind, title, description, sections, …} → attach_artifact as Preview — <title>. Instagram: one Instagram-style viewer over the slide asset URLs, never one artifact per slide.
4. Check — a separate session runs blog-checker: content_lint, preview_render, judgement and fact checks, then check_record {tenant_slug, subject_kind: "article" | "page" | "instagram", subject_ref: <issue id>, round, verdict, findings, models}. On pass it raises the Context request_review; on bounce the maker fixes and re-upserts with the same id.
5. Approve — the owner decides in Context. Locale variants are child Issues of the EN master and inherit its approval (cascade) once their own verdict is attached.
6. Publish — only for an Issue whose Context state is done: article_set_status {tenant_slug, id, status} forward as needed, then publish {tenant_slug, assert_context_done: true, context_issue_id, article_id | page_id | instagram_post_id}. Publish at the slot (due), never early; a missed window is re-slotted, never silently published late. Post the live URL on the Issue, then complete_issues with work_stats.
7. Measure — stats_summary, top_posts, post_stats, ai_referrals, search_queries, indexing_status; the weekly blog-assessment probes AI visibility (visibility_queries_set → probe → visibility_results_ingest → visibility_report) and posts one report Issue.

6. Rules that bite

• Never publish without a Context approval. publish requires assert_context_done: true and refuses when the Issue is not done. Never mark that Issue done yourself.
• Maker and checker are different sessions; a checker run gets only the brief and the draft. No request_review before the verdict, deliverable, preview and models are on the Issue.
• Status only moves forward one step; any other jump is an error. retired is terminal.
• context_issue_id on every upsert — the server rejects content without it.
• No invented facts, numbers, quotes or anecdotes; cite sources in draft notes. No medical, legal or financial claims the owner did not approve.
• Keys by name only. Binary bytes never go through the model: use the upload plan.
• Context calls from this workflow follow the Context guide: caller on every call, expected_version on updates, ≤ 2 Issues per create_issues today.

7. Skills to load

Skill	Role	Raw
rules-blog	the Context Blog harness: epic shape, labels, checker gate, publish gate	https://agents.onecontext.me/skills/rules-blog.md
blog-agent	set-up workflow: interview, chart the tenant Epic, install routines	https://agents.onecontext.me/skills/blog-agent.md
blog-drafter / blog-checker / blog-publisher	nightly maker, independent checker, 3-hourly publisher	https://agents.onecontext.me/skills/blog-drafter.md · blog-checker.md · blog-publisher.md
instagram-drafter / instagram-publisher	carousel maker and scheduler	https://agents.onecontext.me/skills/instagram-drafter.md · instagram-publisher.md
site-builder	site pages workflow	https://agents.onecontext.me/skills/site-builder.md
blog-assessment / daily-brief	weekly report; daily heartbeat	https://agents.onecontext.me/skills/blog-assessment.md · daily-brief.md
rules, setup-context	Context harness (see the Context guide)	https://agents.onecontext.me/skills/rules.md

Channel rule: setup serves the tenant's channel (latest unless the tenant opted into beta). Routines refresh their .claude/skills overlay from the channel before each run and record the versions in the run log; if what you hold differs from the channel and you cannot refresh, finish, comment on the Issue, and tell the owner.

8. Tool map

Links: https://agents.onecontext.me/mcp/context-blog#<tool>.

Stage	Tools	Notes
Skills & setup	usage_guide, get_capabilities, list_skill_catalog, get_skill, setup	all read-only; setup first
Tenant & brand	tenant_create, tenant_get, tenant_update, brand_upsert, audiences_upsert, hubs_upsert, topics_upsert, topics_list	upserts replace by key/slug
Content	article_upsert, article_get, article_list, article_set_status, page_upsert, page_get, page_list, page_set_status, publish	publish takes exactly one id
Checker	content_lint, preview_render, check_record, check_list	lint is read-only and side-effect free
Assets	asset_upload, asset_complete	signed PUT, then complete with width/height/alt
Instagram	instagram_post_upsert, instagram_post_get, instagram_post_list, instagram_post_set_status	slides are asset ids; caption ≤ 2200 chars
Domain	domain_connect, domain_status	hosted tier; DNS/proxy instructions
Analytics	stats_summary, top_posts, post_stats, ai_referrals, search_queries, indexing_status, tracking_setup_status	periods `7d
AI visibility	visibility_queries_set, visibility_results_ingest, visibility_report	probe runs agent-side with the host's keys

9. Errors and limits

• Lint failures come back as findings with a path; article_upsert refuses until they are fixed — do not work around the lint.
• Status errors — moving more than one step, or publish on a row that is not approved, or assert_context_done false: fix the state, do not retry blindly.
• VALIDATION_FAILED — a required field is missing (tenant_slug, context_issue_id, translation_group, hub_slug are the usual ones).
• HTTP 410 {"error":"moved","endpoint":…} — retired host; switch to the endpoint in the body.
• Assets: use the signed upload URL before it expires; complete the row or it is not referenceable.
• Approvals, review requests and every error class of the Context side (PAIRING_REQUIRED, OPERATION_PENDING, WORKER_RESOURCE_LIMIT/546 with backoff, 25 MB artifact cap, 25 s bounded wait) apply unchanged when this workflow writes to Context.
• Rate: one new EN piece per drafter run plus its locale variants; one new carousel per run; publisher runs every 3 h and publishes only inside [due, due + 3 h).