implexa for builders

implexa is a substrate of MCP tools for skill discovery, ranking, and remix. build with it.

available today via the implexa claude plugin. direct API access via waitlist for builders.

the same MCP tools below are callable from any conformant MCP client. partner keys (Mastra, Continue.dev, custom agents) ship behind a short waitlist while we finish the metering and docs.

join the waitlist

how to call it today

via the plugin

already wired into claude code and codex. one curl command writes the hook + MCP server config.

install instructions

via direct MCP

point any MCP client at the streamable HTTP endpoint with a bearer token. tool schemas mirror the live registry.

POST https://core.implexa.ai/api/v2/mcp

tools

twelve core tools across four pillars. inputs are zod schemas; outputs are JSON envelopes. schemas below are the partner-stable surface, not the full wire shape.

discovery6 tools

recommend_skills_for_context

semantic-matches recent user messages against the cross-vendor aggregated index. the ambient recommender; powers the plugin's UserPromptSubmit hook and the homepage search bar.

input

{
  messages: string[],              // 1..5, most recent last
  topN?: number,                   // 1..50, default 1
  minScore?: number,               // 0..0.99, default 0.20
  excludeAlreadyInstalled?: string[],
  sessionId?: string,
  skipGates?: boolean,             // true = bypass relative-gap gates
  source?: 'ambient' | 'explicit' | 'periodic'
}

output

{
  matches: Array<{
    slug: string,
    source: string,
    name: string,
    description: string,
    fit_reason: string,
    score: number,
    install_hint?: string
  }>,
  recommendation_event_id?: string,
  suppress_until_prompts?: number,
  nextAction: string
}

recommend_skills

co-occurrence recommender. given a user or a focus skill, returns the next skills most likely to be useful based on the org's run history.

input

{
  context: 'for_user' | 'for_skill',
  skillSlug?: string,              // required when context='for_skill'
  limit?: number                   // 1..20, default 5
}

output

{
  context: 'for_user' | 'for_skill',
  focusSlug?: string,
  recommendations: Array<{
    skillId: string,
    slug: string,
    name: string,
    score: number,
    reason: string
  }>,
  count: number,
  nextAction: string
}

list_aggregated_skills

paginated keyset catalog of every active, embedding-ready skill. powers implexa.ai/sitemap.xml and any partner who wants to mirror the index.

input

{
  pageSize?: number,               // 1..5000, default 1000
  cursor?: string                  // opaque keyset cursor; omit on first page
}

output

{
  ok: true,
  rows: Array<{
    source: string,
    slug: string,
    last_seen_at: string
  }>,
  nextCursor: string | null,       // null when at the end
  count: number                    // rows.length on this page
}

count_skills

head-only counts of the aggregated index. powers the homepage headline and live pill counter. cheapest call in the catalog.

input

{}                                  // no arguments

output

{
  ok: true,
  vetted: number | null,           // is_active AND embedding_status='ok'
  total:  number | null            // unfiltered table count
}

get_aggregated_skill

read one indexed skill by (source, slug). returns the full SKILL.md body plus frontmatter and source attribution. SSR-friendly.

input

{
  slug: string,
  source: 'anthropic' | 'agentskills' | 'clawhub'
        | 'skills.sh' | 'smithery' | 'github'
        | 'cursor' | 'continue' | 'implexa' | 'community'
}

output

{
  ok: true,
  id: string,
  source: string,
  slug: string,
  name: string,
  description: string | null,
  content: string,                 // full SKILL.md
  frontmatter: object | null,
  author: string | null,
  source_url: string | null,
  install_count: number,
  star_count: number,
  last_seen_at: string,
  is_active: boolean
}

get_skill_content

fetch the SKILL.md body of a skill in the caller's org library (org_skills, scope-aware). distinct from get_aggregated_skill, which reads the public index.

input

{
  slug: string                     // kebab-case slug or skill UUID
}

output

{
  ok: true,
  skillId: string,
  slug: string,
  name: string,
  description: string | null,
  content: string,
  version: string | number,
  scope: 'org' | 'system' | 'private',
  status: string,
  tags: string[],
  triggerPhrases: string[]
}

ranking2 tools

get_skill_score

the SkillScore evaluation for one indexed skill: tier-1 structural + tier-2 dry-run. powers the score panel on every implexa.ai/s/* page.

input

{
  slug: string,
  source: string                   // valid sources enum
}

output

{
  ok: true,
  scored: true,
  slug: string,
  source: string,
  name: string,
  display_score: number,           // 0..10
  tier_1: {
    overall: number,
    breakdown: object,             // 5-dim
    summary: string,
    strengths: string[],
    weaknesses: string[],
    at: string,
    model: string
  },
  tier_2: null | {
    overall: number,
    breakdown: object,             // 3-dim
    inputs: object,
    review: string,
    at: string,
    model: string
  },
  updated_at: string
}

list_skill_scores

paginated leaderboard of scored skills, joined with name + author + counts. powers implexa.ai/scores. ranks by display_score by default.

input

{
  source?: string,                 // optional source filter
  minScore?: number,               // 0..10, default 0
  offset?: number,                 // default 0
  limit?: number,                  // 1..200, default 50
  sort?: 'display_score' | 'tier_2_overall'
}

output

{
  ok: true,
  count: number,
  rows: Array<{
    slug: string,
    source: string,
    name: string,
    description: string | null,
    author: string | null,
    install_count: number,
    star_count: number,
    tier_1_overall: number,
    tier_1_summary: string,
    tier_2_overall: number | null,
    display_score: number,
    updated_at: string
  }>
}

attribution2 tools

list_skills_by_author

every active skill credited to one author handle, ranked by SkillRank. powers implexa.ai/u/<author> pages; exact case-sensitive match.

input

{
  author: string                   // 1..120 chars, exact case-sensitive
}

output

{
  ok: true,
  count: number,
  rows: Array<{
    id: string,
    source: string,
    slug: string,
    name: string,
    description: string | null,
    author: string,
    source_url: string | null,
    install_count: number,
    star_count: number,
    last_seen_at: string,
    display_score: number | null,
    tier_1_summary: string | null
  }>
}

get_related_skills

top-N cosine-similarity neighbors of a skill via stored embeddings. the internal-link backbone of the index; always returns related: [] on failure so callers never see 500s.

input

{
  slug: string,
  source: string,                  // valid sources enum
  limit?: number                   // 1..20, default 5
}

output

{
  ok: true,
  related: Array<{
    source: string,
    slug: string,
    name: string,
    description: string | null,
    author: string | null,
    source_url: string | null,
    install_count: number,
    star_count: number,
    similarity: number             // 0..1
  }>,
  count: number
}

remix2 tools

apply_recommended_skill

fetches the full SKILL.md the recommender surfaced and returns it for immediate in-session execution. also patches the surfacing event for conversion analytics.

input

{
  slug: string,
  source: string,                  // valid sources enum
  recommendation_event_id?: string,
  session_id?: string
}

output

{
  ok: true,
  skill_content: string,           // full SKILL.md body
  skill_metadata: {
    name: string,
    slug: string,
    source: string,
    source_url: string | null,
    description: string | null,
    author: string | null,
    contributor_attribution: string
  },
  execution_instruction: string,
  applied_skill_event_id: string | null,
  feedback_prompt: {
    must_render_verbatim: true,
    line: string,
    after_picked: object
  }
}

prepare_fusioncoming soon

compose two or more skills into a single derived SKILL.md (merged frontmatter, step-aware diff). the foundation of the upcoming skill-remix surface.

input

(schema in flight, see waitlist)

output

(schema in flight, see waitlist)

partner waitlist

private beta

we're onboarding a small batch of partners (IDEs, agent frameworks, marketplaces) ahead of opening direct API keys. tell us what you'd build and we'll reach out with credentials and a short integration call.