API Reference
cortex-engine exposes 17 core MCP tools. These are available via the MCP server or the SDK directly.
Memory Tools
observe(content)
Store a new observation. The content is embedded and stored with an initial salience score.
await cortex.observe("User prefers TypeScript over Python");
query(question)
Semantic search over memories. Returns ranked results by relevance and salience.
const results = await cortex.query("language preferences");
// → [{ content: "User prefers TypeScript...", salience: 0.92 }]
recall(topic)
Retrieve memories related to a specific topic with full context.
forget(id)
Remove a specific memory by ID.
Cognition Tools
validate(claim)
Check a claim against existing memories. Returns supporting and contradicting evidence.
const result = await cortex.validate("User likes Python");
// → { contradicts: true, evidence: [...] }
believe(position)
Record a belief or position. Beliefs are tracked over time and can evolve.
evolve(change, reason)
Propose an identity or belief change with an auditable reason.
wander()
Graph walk through the memory network. Returns unexpected connections and dormant memories.
Thread Tools
thread_create(title)
Start a new persistent thought thread.
thread_update(id, content)
Add progress to an existing thread.
thread_resolve(id, resolution)
Close a thread with a resolution.
Operational Tools
ops_append(entry)
Log an operational breadcrumb.
journal_write(reflection)
Write a session reflection.
vitals_get()
Read current agent vitals (energy, focus, emotional state).
neighbors(concept)
Get connected nodes in the memory graph.
Configuration
All tools respect the namespace of the current agent. Storage provider and embedding engine are set in cortex.config.ts.
- Getting Started — Quick setup guide
- Architecture — How it all fits together
- GitHub — Source code
- npm — Package