Twitter/X — Dual-Mode Integration

Three tools for different operations. Choose by task:

Setup

x-search (Official API v2)

Requires an X API Bearer token from console.x.com. Pay-per-use — no monthly commitment.

# Store token (pick one)
echo 'X_BEARER_TOKEN=your_token_here' >> ~/.claude/skills/twitter/.env
# OR
export X_BEARER_TOKEN=your_token_here
# OR
mkdir -p ~/.config/env && echo 'X_BEARER_TOKEN=your_token_here' >> ~/.config/env/global.env

OAuth 1.0a (for posting)

Posting requires OAuth 1.0a credentials in addition to the Bearer token.

1. Go to console.x.com → your project → Keys and tokens
2. Under "Consumer Keys", copy API Key and API Key Secret
3. Under "Authentication Tokens", generate Access Token and Access Token Secret (with Read and Write permissions)
4. Add all four to ~/.claude/skills/twitter/.env:

X_API_KEY=your_api_key
X_API_SECRET=your_api_secret
X_ACCESS_TOKEN=your_access_token
X_ACCESS_TOKEN_SECRET=your_access_token_secret

bird CLI

brew install steipete/tap/bird   # or: npm install -g @steipete/bird
bird whoami                       # verify auth (uses browser cookies)

Smaug (archival)

cd ~/tools/smaug && npx smaug setup

x-search — Official API Research

Run via: bun run ~/.claude/skills/twitter/x-search/x-search.ts <command>

Quick Search (cost-conscious default)

bun run ~/.claude/skills/twitter/x-search/x-search.ts search "AI agents" --quick
# 1 page, max 10 results, auto noise filter, 1hr cache, cost summary
# Est. cost: ~$0.50

Full Search

bun run ~/.claude/skills/twitter/x-search/x-search.ts search "Claude Code" --pages 3 --sort likes --since 1d
bun run ~/.claude/skills/twitter/x-search/x-search.ts search "BG3 mods" --from LarianStudios --quality
bun run ~/.claude/skills/twitter/x-search/x-search.ts search "Minoan archaeology" --markdown --save

Search Options

Profiles, Threads, Single Tweets

bun run ~/.claude/skills/twitter/x-search/x-search.ts profile AnthropicAI --count 10
bun run ~/.claude/skills/twitter/x-search/x-search.ts thread 1234567890 --pages 3
bun run ~/.claude/skills/twitter/x-search/x-search.ts tweet 1234567890

Watchlist

Monitor accounts with batch checking.

bun run ~/.claude/skills/twitter/x-search/x-search.ts watchlist                    # show list
bun run ~/.claude/skills/twitter/x-search/x-search.ts watchlist add kikismith "Ancient art"
bun run ~/.claude/skills/twitter/x-search/x-search.ts watchlist remove kikismith
bun run ~/.claude/skills/twitter/x-search/x-search.ts watchlist check              # check all accounts

Watchlist stored at ~/.claude/skills/twitter/x-search/data/watchlist.json.

Posting and Replying (OAuth 1.0a)

Post tweets and replies via the official API. Requires OAuth 1.0a credentials (see setup above).

# Post a tweet
bun run ~/.claude/skills/twitter/x-search/x-search.ts post "Hello from x-search!"

# Reply to a tweet
bun run ~/.claude/skills/twitter/x-search/x-search.ts reply 1234567890 "Great thread!"

• 280-character limit enforced
• Cost: $0.01 per post
• Returns the posted tweet URL
• Errors: descriptive messages for missing credentials, permission issues, rate limits

Cache

15-minute TTL (1hr in quick mode). File-based, auto-managed.

bun run ~/.claude/skills/twitter/x-search/x-search.ts cache clear

Cost Reference

X API deduplicates reads within 24h. Cache prevents redundant queries within 15min.

bird CLI — Session-Based Operations

Uses undocumented Twitter GraphQL APIs via browser cookies. Free but may break when Twitter rotates internals.

Posting

bird tweet "Hello world"
bird reply 123456789 "Great thread!"
bird tweet "With image" --media photo.png --alt "Description"

Reading

bird read 123456789
bird thread https://x.com/user/status/123456789
bird replies 123456789 --json

Search (free, via GraphQL)

bird search "query" -n 10
bird search "from:anthropic" -n 20 --json

Monitoring

bird mentions -n 10
bird bookmarks -n 20
bird likes -n 20
bird following -n 50
bird followers -n 50

Maintenance

bird whoami          # verify auth
bird check           # credential sources
bird query-ids --fresh  # refresh when Twitter rotates IDs

Smaug — Bookmark Archival

Archives bookmarks/likes to organized markdown with AI categorization.

cd ~/tools/smaug
npx smaug run                    # fetch + process with Claude
npx smaug fetch 20               # fetch only
npx smaug fetch --source likes   # likes instead of bookmarks
npx smaug run -t                 # track token costs
npx smaug status                 # check state

Output: bookmarks.md, knowledge/tools/, knowledge/articles/

Research Methodology

For deep X research, follow this agentic loop:

1. Decompose — Break topic into 2-3 search queries targeting different angles
2. Search — Run each query with --quick first to assess signal quality
3. Refine — Narrow queries based on initial results (add --from, --quality, --since)
4. Thread — Follow high-value conversation threads for deeper context
5. Deep-Dive — Re-run best queries with --pages 3 --markdown --save
6. Synthesize — Compile findings from saved markdown into analysis

Default to --quick for exploratory queries. Use full mode only for confirmed-valuable searches.

Troubleshooting

x-search: "X_BEARER_TOKEN not found"

Set token via env, ~/.config/env/global.env, or ~/.claude/skills/twitter/.env. Get one at console.x.com.

bird: 403 Errors

Browser cookies expired. Log into Twitter in browser, then bird check.

bird: Query ID Errors

Twitter rotated internals. Run bird query-ids --fresh.

Rate Limiting (429)

Wait for reset. x-search shows reset time. bird: wait a few minutes.

When to Use Which