Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/objectstack-ai/claude-plugin/connectgit clone --depth 1 https://github.com/objectstack-ai/claude-pluginWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00013 | $0.00503 |
| Opus 5 | $0.00006 | $0.00251 |
| Sonnet 5 | $0.00003 | $0.00101 |
| Haiku 4.5 | $0.00001 | $0.00050 |
Grade A, and why
connect scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Connect the user to an ObjectStack deployment over MCP. Follow these steps exactly:
-
Determine the MCP URL.
- If the user provided an argument, use it:
$ARGUMENTS. - Otherwise ask for their deployment URL (the address they open in the browser, e.g.
https://acme.objectos.appor an intranet host). Their admin can also find the exact MCP URL on the Setup → "Connect an agent" page. - Normalize: if the URL does not already end in
/api/v1/mcp, append/api/v1/mcp. Keep the scheme as given (intranet deployments may behttp://on a loopback/dev host).
- If the user provided an argument, use it:
-
Pick a server name. Default to
objectstack; if the user already has a server with that name (check withclaude mcp list), derive one from the host (e.g.objectstack-acme) so multiple environments can coexist. -
Add the server. Run:
claude mcp add --transport http <name> <url>ObjectStack deployments are their own OAuth 2.1 authorization servers with Dynamic Client Registration — no credentials belong in this config. On first use, the client opens a browser login; the user signs in as themselves and every tool call runs under their own permissions and row-level security.
-
Headless fallback (only if the user asks, e.g. for CI or a machine account). An ObjectStack API key can be sent as a header instead:
claude mcp add --transport http <name> <url> --header "x-api-key: <key>"Warn that the key is a password-equivalent and is shown only once at creation.
-
Verify. Call the server's
list_objectstool once. If it returns objects, tell the user they're connected and mention theobjectstackskill now guides data work. If it returns 401, the OAuth login likely wasn't completed — ask them to retry the tool call and finish the browser login. If the URL is unreachable, confirm the deployment is running and the MCP surface isn't disabled (OS_MCP_SERVER_ENABLED=false).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 32 lines · 13 tokens per session scan A 62ca8eb1e152
connect is a command published in the GitHub repository objectstack-ai/claude-plugin (0 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 13 tokens to every session and 503 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
e2e
使用 Playwright 对 Web UI 进行端到端测试（支持视频录制、Trace 录制、控制台/网络日志捕获）.
auto-browse
Auto-browse — learn, optimize, and graduate browser operations or web data-mining workflows.
research-perplexity
Run a deep research query using Perplexity's /research mode via Playwright browser automation. This is an alternative to /export-to-council that uses Perplexity's dedicated research mode instead of multi-model council.
review
Compare a reference design against an implementation. Accepts Figma URL, image file, or browser URL as reference.
graphify
Turn your vault into a clustered knowledge graph with HTML and JSON outputs.
laravel-playwright
E2E Playwright patterns; use the laravel:e2e-playwright skill exactly as written.