AGENTS.md Tabnine AI coding configuration setup guide enterprise 2026

Does Tabnine Read AGENTS.md? No — It Has Its Own Guidelines Folder, and Admin Can Override Yours (2026)

The Prompt Shelf ·

Tabnine Agent does not read AGENTS.md, CLAUDE.md, or any other tool’s instruction file. It has its own convention — Markdown files under .tabnine/guidelines/ — and Tabnine’s own documentation describes it explicitly as an analog: “think of these in a similar fashion to the agents.md file that other agentic tools use.” That framing is honest about what it is (a comparable idea) and what it isn’t (compatible). If you’re rolling out Tabnine alongside Codex, Claude Code, or Cursor on the same repo and expect your existing AGENTS.md to just work, it won’t — and there’s a second wrinkle specific to Tabnine’s enterprise setup: an Admin Console can override your personal guidelines file without your local repo changing at all.

Quick answer

QuestionAnswer
Does Tabnine Agent read AGENTS.md?No
Does it read CLAUDE.md?No
What does it read instead?Markdown files in .tabnine/guidelines/ (project) or ~/.tabnine/guidelines/ (personal)
Does Tabnine call this out as similar to AGENTS.md?Yes, in its own docs, but not as a compatibility claim
Can guidelines be overridden without touching the repo?Yes — an Admin Console (org-level) setting takes precedence over your local guidelines
Does the CLI behave differently from the IDE extension?No — per Tabnine’s docs, “guidelines configured in the Admin Console apply to CLI sessions in the same way they apply to IDE chat sessions”
How fast do changes propagate?Up to 15 minutes, or immediately on IDE/extension restart

What actually lives in .tabnine/guidelines/

Per Tabnine’s own documentation, guidelines are Markdown files stored in your project’s .tabnine/guidelines/ directory (or ~/.tabnine/guidelines/ for personal, cross-project defaults) that act as custom system prompts — workflow rules, tool instructions, and team standards injected automatically at the start of every Agent session. The docs recommend a hierarchical structure and cap individual files around 500 lines for clarity:

.tabnine/
  guidelines/
    coding-standards.md
    error-handling.md
    testing.md

A typical file follows a template with headers like “Team Coding Standards,” “Overview,” “Instructions” (with Code Style / Error Handling / Testing subsections), and “Examples” — closer in spirit to a set of focused .cursor/rules/*.mdc files than to one all-encompassing root AGENTS.md.

Tabnine Agent also has a second, related-but-distinct directory for reusable procedures: .tabnine/skills/ (or ~/.tabnine/skills/), configured separately from guidelines and closer to what Claude Code calls Skills — task-triggered, not always-loaded.

The part that surprises teams: Admin Console precedence

This is the detail most “AGENTS.md alternative” comparisons skip, because most tools in this space don’t have it: Tabnine’s enterprise tier lets an organization set guidelines centrally through an Admin Console, and those org-level guidelines override your personal .guidelines.md file automatically, org-wide — not per-repo, not opt-in. The documentation is direct about the scope: changes apply “in the same way” to both CLI sessions and IDE chat sessions, and propagate within 15 minutes or immediately on restart.

Practically, this means a developer’s local .tabnine/guidelines/*.md can be silently superseded by whatever the org admin has configured, with no diff in the repository to explain why the agent’s behavior changed. If you’re debugging “why did Tabnine suddenly stop following my error-handling convention,” check the Admin Console before you check your own files — the repo can be completely unchanged while the effective guidelines shift underneath it.

Making an existing AGENTS.md usable with Tabnine

There’s no native fallback and no config flag to point Tabnine at a root AGENTS.md the way some tools’ custom-context mechanisms allow. The practical workaround is to treat AGENTS.md as your source of truth and mirror the relevant sections into .tabnine/guidelines/:

.tabnine/guidelines/project-context.md
# Project Context

<!-- Mirrored from AGENTS.md — keep in sync manually, Tabnine does not read AGENTS.md directly -->

## Build & Test
...

## Code Style
...

Keep the file under the ~500-line guidance and split by concern (build/test, style, architecture) rather than dumping the entire AGENTS.md verbatim into one file — the documentation’s own example structure organizes by topic for a reason: Agent sessions load these as context on every run, so an oversized single file costs tokens on every task regardless of relevance.

FAQ

Is this likely to change now that Tabnine was acquired by Tricentis (July 2026)? No indication either way in current docs. The acquisition was framed as a strategic enterprise-testing integration rather than a product wind-down — Tabnine’s per-seat pricing and product surface were reported unchanged at announcement — but AGENTS.md interoperability isn’t mentioned in any Tricentis-Tabnine materials as of this writing.

If I put an AGENTS.md in my repo root alongside .tabnine/guidelines/, does Tabnine ignore it completely, or read it as a fallback? Ignores it completely. Tabnine’s context sources are .tabnine/guidelines/, .tabnine/skills/, and whatever you attach manually in a chat session (@file references) — a root AGENTS.md sitting unreferenced has zero effect on Agent behavior.

Does the Admin Console override apply to .tabnine/skills/ too, or only guidelines? Tabnine’s documentation covers Admin Console precedence specifically for guidelines; skills configuration is documented as project- or user-level only, with no mention of a centralized org override in the current docs.

Is there a community feature request open for native AGENTS.md support, the way there is for other tools? Not a prominent public one — Tabnine doesn’t run an open GitHub issue tracker for product feature requests the way open-source-adjacent CLI tools (Qwen Code, Amazon Q Developer CLI) do, so this kind of gap doesn’t surface the same way. Requests would go through Tabnine’s own support/feedback channels instead.


For a broader map of which coding agents read which config file out of the box, see our AGENTS.md vs CLAUDE.md tool support map. If a context file you know is present still isn’t showing up in a session for a different tool, our AGENTS.md not loading debugging guide covers tool-by-tool verification steps. Browse more real-world instruction files and setup patterns in our gallery.

Related Articles

Explore the collection

Browse all AI coding rules — CLAUDE.md, .cursorrules, AGENTS.md, and more.

Browse Rules