Servidor MCP de Flux

• Nome no registroapp.fluxgraph/linkedinverificado no registro MCP

O nome no registro fica sob fluxgraph.app. O registro só libera esse espaço de nomes para quem prova controle do domínio por DNS ou HTTP.

• Estadorespondeverificado em 2026-11-02
• Ferramentas66
• O servidor se chamaflux
• Versão do servidor0.1.0
• Versão do protocolo2025-06-18

Em 2026-11-02 o servidor respondeu a initialize e tools/list sem login. Primeira resposta em 107 ms.

• As ferramentas fazem todos os casos de uso verificados: Compare which post topics and formats earn the most engagement; Find the best days and hours to publish; Score a draft post before publishing it.
• Verificado em 2 dos 7 dias necessários para julgar a confiabilidade.

Dito por regras a partir das medições acima. A primeira chamada provável é a escolha de um modelo na lista de ferramentas; leia a descrição antes de confiar nela.

• Nota INITE MCPdepois de uma semanaa nota precisa de 7 verificações diárias: 2 até agora
• O que um agente consegue fazer com ele3 de 3 casos de uso
• Respondeumedindo2 das 7 verificações diárias necessárias para mostrar a porcentagem
• Mediana da primeira resposta142 ms
• Espaço no contexto do agente≈ 14.719 tokensas definições de todas as ferramentas (67), que o agente carrega a cada turno
• Definições das ferramentasB · 83 de 100
• Resultados tipados0 de 67 ferramentasdeclaram outputSchema: o resultado chega tipado e é lido de uma vez

• Sem dizer se só lê ou se pode destruir: 67 de 67

Um programa verifica o servidor uma vez por dia: initialize e tools/list, sem modelo. Os tokens são o tamanho das definições, a cerca de quatro caracteres por token. A nota é a média de seis verificações do nome, da descrição, do esquema e das anotações de cada ferramenta; diz se as ferramentas estão bem descritas, não se funcionam. A nota pesa só o que foi medido: se responde (25%), se faz o trabalho do produto (25%), definições (20%), contexto (10%), uso (10%) e a facilidade de conectar (10%). Uma parte não medida fica de fora, e uma instrução oculta limita a nota a 40. O estado dos servidores MCP dos fornecedores: método e números →

• list_profiles

  List all tracked LinkedIn profiles for the authenticated user. Returns profile IDs needed by other tools.

• preview_profile

  Analyze a public LinkedIn profile or company page without a Flux account: pulls the latest ~50 posts (once per profile) and returns a preview id, post count and a one-line posting summary. Use preview_posting_insights with the preview_id…

• preview_posting_insights

  Best day, hour, format and length for a previewed profile (from preview_profile). No account needed.

• claim_preview

  Make a profile from preview_profile your own Flux dashboard. Requires sign-in; creates your own_profile stream on the free plan pointing at the same posts. Refuses if you already track that profile.

• get_posts

  List a tracked profile's persisted posts with full metrics and text — query against the local post library, no LinkedIn pull, free. Use this when the caller already has a profile_id and wants to read its posts. Filter by date range, post…

• get_profile_summary

  Get KPI summary for a tracked profile: total posts, average engagement (broken down by likes/comments/shares/conversation rate), posts per week, engagement trend, top format, and posting goal status.

• get_posting_insights

  Get optimal posting patterns for a profile: best day of week, best hour, format performance breakdown, and character length sweet spot. Includes full metric breakdown (likes, comments, shares, conversation rate) for every bucket. All…

• get_topic_insights

  Surface the most-frequent topics in a profile's posts, with engagement context. Returns top N unigrams and (optionally) top N bigrams ranked by post count, each with avg_engagement, vs_baseline, and a sample_post_url. Mirrors the…

• get_engagement_trends

  Get engagement time series for a profile with gap-filled buckets, rolling averages, and trend direction. Each bucket includes avg engagement, likes, comments, shares, and conversation rate. Useful for understanding whether engagement is…

• search_posts

  Keyword + semantic search over persisted posts in the local library — full-text + pgvector cosine, no LinkedIn pull, free. Use this when the caller has a query string and wants ranked persisted-post results. Pass `scope` (a…

• get_credit_balance

  Return the authenticated user's current credit balance, this period's monthly allocation, and credits used since the period started. Free.

• get_usage_history

  Return recent credit_ledger transactions for the authenticated user. Filter by tool name, reason, or date range. Paginated. Free.

• get_cost_estimate

  Given an MCP tool name, return its credit cost so you can preview an action before running it. Also accepts a stream type (external_profile, company, topic, person_interests, hiring_signals, ad_watch) to preview the per-page cost…

• create_stream

  Create a new research stream (external profile, company, topic, person interests, hiring signals, or ad watch). Running the stream on its schedule costs credits (see get_cost_estimate for the per-type cost). own_profile streams are not…

• list_streams

  List all research streams for the authenticated user (external profiles, companies, topics, person interests, hiring signals, ad watches, and own_profile). Pass `type` to filter. Replaces list_profiles for the streams-first world, but…

• get_stream_feed

  Return the most recent posts a stream has collected into the library, newest first by published_at. Use this to browse what a topic / external profile / person_interests / company stream has surfaced on its schedule.

• pause_stream

  Pause a stream so it stops running on its cron schedule. Existing posts remain in the library. Use resume_stream to re-enable it. own_profile streams are billing-managed and cannot be paused here.

• resume_stream

  Resume a paused stream. Sets active=true and queues it to run on the next cron tick by clearing next_run_at. own_profile streams are billing-managed and cannot be resumed here.

• delete_stream

  Delete a stream and its stream_runs. Posts collected by the stream stay in your library (stream_id is set to NULL via ON DELETE SET NULL) so historical data isn't lost. Auto-collection is deleted with the stream. own_profile streams…

• run_stream_now

  Trigger an immediate run of an active stream, bypassing its schedule. The run is fire-and-forget — the response confirms the dispatch but post collection happens asynchronously. Credits are charged for pages actually walked, not pages…

• linkedin_profile_info

  Fetch a LinkedIn person profile (headline, current role, work history, education, location). Costs 5 credits per call. Results are upserted into `people` (structured fields) and `linkedin_profiles` (so profile-brief renders) — same…

• linkedin_profile_posts

  Fetch recent posts from a LinkedIn person profile. 5 credits per page of 50, up to 250 posts, charged only for pages actually fetched. Results are stored in the post library (source_type='external') and linked to a linkedin_profiles row…

• linkedin_company_posts

  Fetch recent posts from a LinkedIn company page. 5 credits per page of 50, up to 250 posts, charged only for pages actually fetched. Results are stored in the post library (source_type='external') and linked to a linkedin_profiles row…

• linkedin_post_search

  Search LinkedIn posts by keyword. Costs 5 credits per call. Returns a single page of matching posts.

• linkedin_profile_reactions

  Fetch the recent posts a LinkedIn person has reacted to. Useful for inferring interests. Costs 5 credits per call.

• linkedin_profile_comments

  Fetch the recent posts a LinkedIn person has commented on. Useful for inferring interests and conversation partners. Costs 5 credits per call.

• linkedin_job_search

  Search LinkedIn job postings. Useful for tracking hiring signals at target companies. Costs 5 credits per call.

• linkedin_ad_search

  Search the LinkedIn ad library. Useful for tracking competitor ad creative and messaging. Costs 10 credits per call.

• linkedin_people_search

  Search LinkedIn people. Filter by company, title, location, or free-text keywords. Costs 5 credits per call. Returns a single page of matching profiles.

• linkedin_company_search

  Search LinkedIn companies. Filter by keywords, industryId, or location. Costs 5 credits per call. Returns a single page of matching companies.

• linkedin_post_reactions

  Fetch the people who reacted to a single LinkedIn post. Costs 5 credits per call. Engagers (with reaction kind) are persisted to the post_engagements graph so subsequent queries about who-engages-with-whom are SQL, not API calls.

• linkedin_post_comments

  Fetch the people who commented on a single LinkedIn post (and their comment text). Costs 5 credits per call. Each comment is persisted to post_engagements with the commenter and full comment text.

• find_similar_posts

  Find persisted posts whose embeddings are nearest to a free-text query — pgvector cosine over the local post library, no LinkedIn pull, free. Use when you have a *block of text* (a draft, a paragraph, a competitor's post) and want similar…

• score_draft

  Predict how a draft post will perform against the user's historical data. Uses an OLS regression model with detrending, ridge regularization, and recency weighting — the same model used on the /score page. Returns a predicted engagement…

• suggest_angles

  Given a topic, suggest distinct content angles informed by what performs well in the user's scoped data. Pulls top-performing historical posts as context and calls OpenAI to generate 5 concrete angles. Costs 5 credits per call. Pass…

• list_segments

  List all audience segments for the authenticated user, with member counts. Audience segments are named groups of people (e.g. 'ICP — VPs of Marketing', 'Champions'). Free — costs 0 credits.

• create_segment

  Create a new audience segment (named group of people) for the authenticated user. Set type='manual' for a hand-curated list (use add_to_segment to populate it), type='smart' with a `config` predicate (e.g. {title_contains:'VP'}) to define…

• add_to_segment

  Add one or more people (by person_id) to an audience segment. Existing members are silently ignored. Free — costs 0 credits.

• remove_from_segment

  Remove one or more people (by person_id) from an audience segment. Missing members are silently ignored. Free — costs 0 credits.

• resolve_segment

  Resolve an audience segment into a concrete list of person_ids. For manual / auto_stream segments, returns stored members. For smart segments, evaluates the segment's `config` predicate against the people table and returns matches. Free…

• list_segment_templates

  List available smart-segment templates. Pass a template id to create_segment to pre-fill its `config` predicate without crafting one from scratch. Free — costs 0 credits.

• get_alerts

  Return recent alerts for the authenticated user's streams. An alert is emitted whenever a scheduled stream run surfaces new items. Pass `unread_only=true` to skip alerts the user has already marked read. Free — costs 0 credits.

• create_collection

  Create a new post collection for the authenticated user. Set type='manual' to hand-curate posts with add_to_collection, or type='smart' with a `config` predicate (e.g. {format:'video',min_likes:50}) to define a dynamic collection resolved…

• add_to_collection

  Pin one or more posts (by post_id) into a collection you own — works on any collection type. On a 'manual' collection this is the only membership mechanism; on a 'smart' collection it adds the posts as extras alongside whatever the rule…

• remove_from_collection

  Remove one or more posts (by post_id) from a collection. Missing members are silently ignored. Free — costs 0 credits.

• list_collections

  List all post collections for the authenticated user, with member counts. Includes manual collections, smart collections (predicate-based), and auto_stream collections (system-managed). Free — costs 0 credits.

• resolve_collection

  Resolve a post collection into a concrete list of post_ids. For manual / auto_stream collections, returns stored members. For smart collections, evaluates the collection's `config` predicate against the user's posts and returns matches…

• create_campaign

  Create a Constellation outreach campaign. A campaign groups target accounts, the people you care about at those accounts, and the warm connections you could approach to reach them. Free — costs 0 credits.

• add_campaign_accounts

  Add one or more target companies to a Constellation campaign. Each account is a company you want to reach; map_constellation is then run per-account. Free — costs 0 credits.

• map_constellation

  Map the professional graph around a target company. Finds senior decision-makers at the account, pulls each target's recent posts/reactions/comments to identify engagers, then scores every connection and produces a prioritized outreach…

• score_connections

  Re-score all constellation entries in a campaign and rewrite queue positions from highest to lowest composite score. Use after new engagers or targets are added. Free — costs 0 credits. Note: this tool requires a `campaign_id` because it…

• get_outreach_queue

  Return a prioritized outreach queue. Provide *exactly one* of three scopes: • `campaign_id` — full Constellation queue with composite_score, pool, recommended_action, and prior outreach_log entries (the original campaign-scoped behavior)…

• get_engagement_brief

  For one person, return their recent posts most relevant to what you post about — plus suggested talking points. Use before a warm comment or DM. Free — costs 0 credits. Identify the person via *exactly one* of: `target_id` (a…

• log_outcome

  Append an entry to the outreach log for a Constellation campaign. Use to record connection requests, acceptances, replies, or no-responses so get_campaign_stats can compute acceptance / reply rates over time. Free — costs 0 credits.

• get_campaign_stats

  Summary stats for a Constellation campaign: target count, entry count, outreach volume, and acceptance / reply rates broken down by pool (coworker / engager / mutual) and composite-score tier. Free — costs 0 credits.

• save_post

  Save a LinkedIn post to your library by its URL. Resolves the post and stores it with source_type='saved'. Optionally pins the post into any collection you own (manual or smart — pinning into a smart collection adds it as an extra…

• get_post_engagers_by_company

  Given a LinkedIn post URL or id, return the companies whose people engaged (reacted/commented), ranked by engager count. Joins through person_positions(is_current=true). Pure DB read — no LinkedIn pull, no credits charged. Engagers…

• get_engagement_industry_breakdown

  Histogram of engagers' companies grouped by industry. Pivots on companies.industry_hierarchy (e.g. 'Tech > Software Development'). Depth=1 buckets by top-level, =2 by top+mid, =3 by full hierarchy. Engagers with no current company / no…

• get_alumni_engagers

  Given a company URL or id, return engagers (across your collected posts) who have a position at that company. Relation filter: 'coworker' = current employees, 'alumni' = any past or current position (most useful for page_type='SCHOOL'…

• get_engager_position_history

  Return a person's full employment timeline from the position graph. Person must have engaged on at least one of your collected posts (engagement gate). Sorted current-first, then most-recent past by started_at desc. Pure DB read — no…

As primeiras 60 de 66, na ordem do servidor.