Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/virsanghavi/axis/llms.txt

Use this file to discover all available pages before exploring further.

Before an agent creates a new file or refactors existing code, it should know what already exists. Axis search gives every agent in a project the same up-to-date view of the codebase — ranked hits with file and line numbers, related files that historically change together, and symbol definitions for the top matches. The free tier uses ripgrep with a keyword ranker and needs no index. The Pro tier blends in hosted vector, full-text, and trigram results fused and reranked by an LLM, with a multi-hop deep_search that reads across files and returns a cited answer.
Free tier users get search_codebase with local ripgrep — instant results, no index needed, and no setup. Pro tier blends hosted vector results in when they arrive within budget. An upgrade prompt appears when a free org reaches for hosted search features.

​
Search tiers

Local (free)

search_codebase answers from ripgrep plus a keyword ranker. No index required. Instant. Available to every user.

Hosted Pro

Vector + full-text + trigram retrieval, fused and LLM-reranked, with related files and definitions enrichment. deep_search for multi-hop agentic answers.

​
search_codebase

The primary discovery tool — use it before creating files, before refactoring, and any time you need to find where something lives in the codebase.
​
query
string
required
The search query. Can be a symbol name, a natural-language description of what you’re looking for, or a pattern.
​
projectName
string
The project to search. Optional — defaults to the auto-detected project.
Returns ranked hits with file:line references, plus (on the Pro tier):
  • related — files that historically change together with the top hit. If you’re about to refactor src/auth.ts, these are the files most likely to also need changes.
  • definitions — symbol definitions for what the top hit calls, so you don’t have to chase imports manually.
// Example response
{
  "hits": [
    { "file": "src/auth.ts", "line": 42, "snippet": "export function issueToken(userId: string, role: string) {" },
    { "file": "src/middleware/auth-guard.ts", "line": 12, "snippet": "const token = await issueToken(user.id, user.role);" }
  ],
  "related": ["src/types/token.ts", "tests/auth.test.ts"],
  "definitions": {
    "issueToken": { "file": "src/auth.ts", "line": 42 }
  }
}

​
deep_search — agentic answer engine

deep_search is for “how does X work?” and “where is Y handled, and why?” questions. It reads across multiple files over several hops and returns a cited answer pointing to the exact source lines that support each claim.
​
query
string
required
A natural-language question about the codebase.
​
projectName
string
The project to search. Optional.
deep_search is hosted Pro only — it carries real embedding and LLM cost. Local users get search_codebase’s instant local fallback for all discovery work.

​
Indexing

Search only returns what’s been indexed. The index is incremental and content-hashed: re-runs skip unchanged files, so it’s cheap to call after making changes.

​
index_codebase

Build or refresh the search index for a project.
​
files
array
required
Array of { path, content } objects. Agents pass the files they edited; for a full initial index, use the axis index CLI instead.
​
prune
boolean
When true, deleted files are removed from the index. Requires allPaths to identify which paths should be kept.
​
allPaths
string[]
Complete list of current file paths in the project. Used with prune: true to drop files that no longer exist.
​
projectName
string
The project to index. Optional — defaults to the current workspace.

​
index_file

Index a single file incrementally. Available on the local server only — it reads the file content directly from disk.
​
filePath
string
required
Path to the file to index. The local server reads its content from disk; no content argument needed.
Agents should call index_file after writing a file to keep the index current within the same session. This ensures subsequent search_codebase calls by any agent on the team reflect the latest changes.

​
search_docs

Search indexed documentation — distinct from codebase search. Local server only; backed by the local RAG fallback over indexed docs.
​
query
string
required
The search query.

​
Initial indexing workflow

1

Use the CLI for large initial indexes

The axis index CLI is faster for indexing an entire repository. It walks the repo, respects .gitignore, and uploads only deltas (content-hashed, unchanged files are skipped):
axis index
Run this once when setting up a project. Re-running it after large changes is also faster than calling index_codebase from an agent for many files.
2

Agents index their own changes

After every file write within a session, the agent calls index_file to update the index immediately. This keeps search results fresh without a full re-index.
3

Periodic full refresh

After large refactors or after merging a significant PR, any agent can call index_codebase with prune: true + allPaths to drop deleted files and bring the index up to date.

​
REST API (hosted only)

The hosted intelligence layer is also accessible directly via the REST API for integrations that don’t use MCP. Embed content — add documents or code snippets to the index: 
curl -X POST https://useaxis.dev/api/v1/embed \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AXIS_API_KEY" \
  -d '{
    "items": [
      {
        "content": "This repo uses Bun",
        "metadata": { "filename": "context.md", "source": "agent-instructions" }
      }
    ]
  }'
Search content — query the hosted index directly: 
curl -X POST https://useaxis.dev/api/v1/search \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AXIS_API_KEY" \
  -d '{
    "query": "What runtime does this project use?",
    "limit": 5,
    "threshold": 0.5
  }'
The threshold parameter (0–1) controls the minimum similarity score for a result to be returned. Lower values return more results with lower confidence; higher values return fewer but more relevant hits.

​
Tool availability by tier

ToolLocal (free)Hosted Pro
search_codebase✅ ripgrep + keyword✅ vector + full-text + trigram, reranked
index_codebase✅✅ incremental, content-hashed
index_file✅ reads from disk—
search_docs✅ local RAG—
deep_search—✅ multi-hop, cited
Call search_codebase before creating any new file or starting any refactor. The most common source of duplicate code in multi-agent projects is agents that each independently implement the same utility — because none of them searched first.

Build docs developers (and LLMs) love

Get started for freeTalk to us