Agents and API

MCP tools

Every tool Noetix gives a connected agent over MCP. Searching and reading the brain, writing pages, following skills, and what is kept safe.

These are the tools an agent gets once it is connected. Every tool runs as you, with exactly your permissions in Noetix.

Your brain

Read and write memory directly. No agent, no LLM in the path.

brain_recall
search the brain and get back what it knows, with sources
brain_status
what's in the brain, and which sources feed it
brain_map
every topic the brain covers, or one topic in depth
brain_entity
one thing and its links, or how two things connect
brain_check
test whether the brain already knows something
brain_changes
what changed recently: pages written or edited, and by whom
brain_authors
the teammates who wrote what the brain knows about a topic
work_blockers
what's stuck in Linear and GitHub, each with the recorded reason
brain_flag_gap
note a question the brain couldn't answer, for the team to fill
brain_gaps
the questions the brain couldn't answer, most asked first
brain_close_gap
mark a gap answered, with the page that answers it
brain_forget
remove something from the brain (two steps, confirmed)

Pages

The documents the brain keeps, edited in place with full history.

page_list
every page by title, most recently changed first
page_read
read one of the brain's pages in full, or a past version of it
page_edit
change a passage of a page, or add to its end (Notes for one-off facts)
page_create
start a new page on a topic
page_history
how a page got here: what changed, when and why
page_restore
roll a page back to a version, or bring back a deleted one

Your skills

How your company does things, written down for agents to follow.

skill_search
find the procedure for a task, or list them all
skill_use
read one procedure, step by step, and follow it
skill_draft
write a procedure down as a draft, for an admin to publish
skill_import
bring a repo's skills, commands and rules in as drafts, for an admin to publish

You

Your profile and what your agents remember about you. Private to you.

profile_read
your profile and memories, read at the start of a session
memory_save
remember a preference, a correction, your role or your work
memory_edit
change a memory that is out of date
memory_forget
delete a memory

What an agent can write

Writes are kept to what an agent should do on its own:

  • A page edit keeps every earlier version, so it can always be undone.
  • brain_forget asks for confirmation before anything is removed.
  • skill_draft and skill_import only write drafts, which an admin publishes. Nothing an agent does over MCP publishes a skill.

Your profile and memories

Your agents read your profile at the start of a session with profile_read, and remember your preferences with memory_save. That is private to you and never goes into the brain.

Calls are counted

Every call an agent makes counts toward your workspace's monthly calls, shown in Settings › Usage. They are warned about once they reach the month's limit, and stopped a little past it, so an agent mid-task isn't cut off the moment you reach it. See Plans and billing.