Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/synpulse8-opensource/pulse8-ai-cortex-knowledge-vault/cortex-mcpnpx skills add synpulse8-opensource/pulse8-ai-cortex-knowledge-vault --skill cortex-mcpgit clone --depth 1 https://github.com/synpulse8-opensource/pulse8-ai-cortex-knowledge-vaultWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/synpulse8-opensource/pulse8-ai-cortex-knowledge-vault/cortex-mcp)<a href="https://agentmods.dev/skills/synpulse8-opensource/pulse8-ai-cortex-knowledge-vault/cortex-mcp"><img src="https://agentmods.dev/badge/skills/synpulse8-opensource/pulse8-ai-cortex-knowledge-vault/cortex-mcp.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00041 | $0.00685 |
| Opus 5 | $0.00020 | $0.00342 |
| Sonnet 5 | $0.00008 | $0.00137 |
| Haiku 4.5 | $0.00004 | $0.00068 |
Grade A, and why
cortex-mcp scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Cortex MCP
Transports
| Mode | How | Config |
|---|---|---|
| HTTP | http://localhost:8420/mcp/ |
CORTEX_MCP_TRANSPORT=http, Docker or scripts/serve.py |
| stdio | Claude Desktop subprocess | CORTEX_MCP_TRANSPORT=stdio |
HTTP uses FastMCP streamable HTTP (stateless_http=True, JSON responses).
Authentication
AUTH_METHOD |
MCP behavior |
|---|---|
none |
Open |
apikey |
x-api-key header must match API_KEY / CORTEX_API_KEY |
oidc |
FastMCP OIDCProxy (browser login); API key fallback if configured |
REST uses AuthMiddleware; MCP uses ApiKeyGuardMiddleware when apikey. See cortex-auth.
HTTP session flow
POST /mcp/—initialize(JSON-RPC 2.0)- Save
mcp-session-idresponse header if present notifications/initializedtools/callwithnameandarguments
Header: Accept: application/json, text/event-stream
Implementation map
| Layer | File |
|---|---|
| Tool handlers (shared with REST logic) | cortex/mcp/tools.py |
| HTTP tool registration | cortex/mcp/http_server.py |
| stdio server | cortex/mcp/server.py |
| Mount on FastAPI app | mount_mcp_on_app() in http_server.py |
Handlers receive injected services: vault_path, graph, qmd, qmd_debounce, compiler.
Cursor integration
This project may register a local MCP server (e.g. user-cortex_local) pointing at http://localhost:8420/mcp/ with x-api-key when auth is enabled.
Use MCP tools for agent workflows; prefer vault_context for RAG-style retrieval.
Tool list
See reference.md for all 10 tools (including
vault_resource_read), parameters, and return shapes.
MCP resources (cortex://resource/{id})
Token-heavy tool outputs can be opted into the server-side resource
store via as_resource: true on vault_search / vault_context
(and GET /api/v1/search?as_resource=true on REST). The store is
in-memory, TTL-bounded, and LRU-capped; tune with
CORTEX_RESOURCE_TTL_SECONDS and CORTEX_RESOURCE_MAX_ITEMS. See
reference.md for details.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 77 lines · 41 tokens per session scan A 084dee9dbfb3
cortex-mcp is a skill published in the GitHub repository synpulse8-opensource/pulse8-ai-cortex-knowledge-vault (14 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 41 tokens to every session and 685 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
session-ingest
Ingest a coding session JSONL into AKB as structured notes — session report + parallel-drafted TIL / task / idea / decision sub-notes. Auto-discovers the current session's JSONL when no positional path is given (Claude / Codex). --delegate {target} fires the work to another target's CLI and exits.
akb-query
Answer a question from the AKB vault — decompose, search (hybrid / graph), ground in compiled-truth-over-timeline precedence, and synthesize a cited answer with gap/conflict/stale flags. Read-only.
llm-wiki-query
Answer operator questions using the LLM Wiki as the primary source of truth, and optionally file reusable syntheses back into the wiki. Use whenever the operator asks a question about wiki content, queries the knowledge base, requests a synthesis across sources, or wants an answer based on what has been ingested.
akb-ingest
Ingest whatever you point at into an AKB vault — a local file, a web URL, a GitHub PR/release/commit, a Confluence page, or a Jira issue. Auto-detects the source type and dispatches to a specialized ingest subagent; the router fetches and writes nothing itself. One target per invocation; globs expand to a sequential…
wiki-write
Add or update wiki content — autonomous ingest from URL, file, or text; autonomous update of existing pages. Auto-creates .wiki/ on first use. Use when: 'save to wiki', 'remember this', 'note this', 'store this', 'add to knowledge base', 'save findings', 'save research', 'save idea', 'write to wiki', 'ingest', 'add…
wiki-view
Wiki dashboard — browse pages, view stats, knowledge graph, export. Use on: 'wiki dashboard', 'wiki stats', 'show wiki', 'wiki pages', 'list wiki', 'wiki graph', 'export wiki', 'wiki overview'.