Querying Instagram from Claude Code: wiring up HikerAPI's MCP server

If you work with social data, you know the drill: to answer one simple
question — "how many followers does this profile have?" or "what are the
top posts for this hashtag?" — you bounce between the Instagram app, a
spreadsheet, and some throwaway script you have to keep alive.

I wanted to ask those questions directly inside my AI assistant (Claude
Code) and get structured data back — without writing and maintaining
scraping code. Here's the setup that solved it: HikerAPI (a read-only
Instagram data API) exposed through MCP (the Model Context Protocol), the
open standard that lets AI assistants call external tools.

What you actually get

The hikerapi-mcp server auto-generates 100+ read tools from HikerAPI's
OpenAPI spec. You ask in natural language; the assistant picks the right
endpoint and hands you structured JSON. No glue code.

Tool group	~Tools	Examples
User Profile	36	get_v2_user_by_username, get_v2_user_by_id
Post Details	20	get_v2_media_info_by_code, ..._comments
Search	13	get_v1_search_users, get_v1_search_hashtags
Hashtags	7	get_v2_hashtag_medias_top / ..._recent
Stories	7	get_v2_story_by_url, get_v1_story_by_id
Location	7	get_v1_location_medias_recent

Prerequisites

• Node.js 18+
• An MCP-capable assistant — Claude Code, Claude Desktop, Cursor, Windsurf, Zed, or OpenAI Codex
• A HikerAPI key from hikerapi.com/tokens

Setup (Claude Code)

The README's quick command works:

claude mcp add hikerapi -e HIKERAPI_KEY=your-key -- npx -y hikerapi-mcp

…but it writes your key in plain text into .claude.json. If you sync or
version your config across machines (I do), that leaks the key.

The cleaner way — store the key in an environment variable and let the
config hold only a pointer to it:

# set the key once in your shell environment
export HIKERAPI_KEY="your-key"          # macOS/Linux
# setx HIKERAPI_KEY "your-key"          # Windows (PowerShell), then reopen

claude mcp add hikerapi --scope user \
  -e 'HIKERAPI_KEY=${HIKERAPI_KEY}' \
  -- npx -y hikerapi-mcp

Now your config literally stores ${HIKERAPI_KEY} — the real secret never
touches the file.

Keep the tool list lean

100+ tools is a lot of surface area for the model's context. Filter to the
groups you actually use with HIKERAPI_TAGS:

claude mcp add hikerapi --scope user \
  -e 'HIKERAPI_KEY=${HIKERAPI_KEY}' \
  -e 'HIKERAPI_TAGS=User Profile,Post Details,Search,Hashtags,Stories' \
  -- npx -y hikerapi-mcp

Validate

Restart the assistant and ask:

Get the Instagram profile for @nasa.

Find the 5 most recent posts for #photography.

If you get structured data back, you're live.

Config reference

Variable	What it does	Required
HIKERAPI_KEY	Your HikerAPI access key	yes
HIKERAPI_URL	Base URL override	no
HIKERAPI_TAGS	Comma-separated groups to include	no
HIKERAPI_EXCLUDE_TAGS	Groups to exclude	no
HIKERAPI_TIMEOUT_MS	Per-request timeout (default 30000)	no
HIKERAPI_MAX_RESPONSE_BYTES	Max response size (default 10 MB)	no

A real use case

For me this became part of an AI marketing workflow: competitor research
(pull follower counts and engagement for a list of handles), hashtag
research before a campaign, and tracking the top posts in a niche — all from
natural-language prompts, without leaving the assistant.

Gotchas

• HikerAPI is prepaid. With a zero balance, endpoints return HTTP 402 Payment Required even though the key itself is valid (auth errors are 401/403). Top up before you validate.
• npx -y downloads the package on first run, so the first startup is slower.
• Use environment variables for the key (see above) — don't commit secrets.

Wrap up

MCP + HikerAPI gets you structured Instagram data inside your AI assistant,
with zero glue code to maintain. Scope the tool tags to what you need, keep
the key in an env var, and you're a couple of commands away from asking your
assistant real questions about real profiles.

Repo: github.com/subzeroid/hikerapi-mcp · Package: npm i hikerapi-mcp