BRING Human expert guidance inside your coding agent

Know your product’s next move with expert guidance.

Connect the CLI and send a focused question, with your app context, to a human expert. The platform spins up a fresh private scratch repo for the review, the expert answers, and you get back an agent-ready prompt straight into your coding agent.

cli mcp guidance expert prompt

Try the CLI Connect MCP

session copy

$ specloop ask --tool codex
daWPmdSWLeptsu84cmejx [guidance] open_for_claim
Query: What should I prompt Codex to do next for auth?
Summary: Added bearer middleware and /me route
URL: https://github.com/acme/shop/pull/42
Claims: 0/1. Responses: 0. Cost: 100 credit(s).

$ specloop get daWPmdSWLeptsu84cmejx
Human expert guidance:
1. Marcel Perro: Prompt Codex to add token expiry validation,
   rate limiting, malformed-header rejection, and
   missing-token tests.

$ npx @specloop/cli ask --tool codex

See it in action

From stuck to directed in 3 steps

A builder asks the community for technical direction. Exact expert comments and a stronger prompt come back where they work.

claude code

1 Trigger request

$ specloop ask --tool claude-code
Query: What should I prompt next for auth?
Summary: Added bearer middleware + /me route.
GitHub: github.com/acme/shop/pull/42
Submitted guidance req_7Qe91

specloop · inbox

2 Expert responding

Collecting expert guidance...

A community reviewer claims and comments 1 expert

A vetted expert guides the builder on context, architecture, prompts, and next steps.

claude code

3 Answer delivered

prompt ready synthesized from expert comments

Expert reply:
Implement this direction with tests and guardrails.
Use expert comments as source-of-truth context.

approved back in your agent

Workflow

Built for builder direction moments

builder-shares-context

The builder sends a focused technical question with approved app context, latest action summary, code snippet, logs, or URL.

expert-responds

The platform queue routes the request to an available community code reviewer, who claims it and provides exact guidance comments.

prompt-returns

The builder sees raw expert comments in the platform and gets an agent-ready prompt through CLI or MCP.

MCP ready

Connect Claude Code, Codex, Base44, Lovable, or other agent tools.

The MCP server exposes tools that an agent can call directly: submit a guidance request, fetch exact expert comments, and retrieve the synthesized agent-ready prompt.

connect mcp copy

SPECLOOP_API_URL=http://localhost:4317 \
SPECLOOP_TOKEN=<builder-token> \
$ npx @specloop/cli mcp

Tools:
- submit_human_guidance
- get_expert_comments
- get_agent_ready_prompt

Credits

Buy credits, spend them on expert guidance

starter

100 credits

$20 · covers 1 request. Good for trying expert direction from a coding agent.

pro

500 credits

$100 · covers 5 requests. For active builders asking experts what to prompt next.

team

1000 credits

$200 · covers 10 requests. For teams using human guidance alongside multiple coding agents.

Run locally

Test the full loop from the terminal

install copy

$ npm i -g @specloop/cli
$ specloop login --email builder@example.com --name Builder --role builder
$ specloop credits buy --pack starter
$ specloop ask --query "What should I prompt next for auth?" --summary "Added bearer auth" --code "app.get('/me', requireUser, handler)"

$ specloop login --email expert@example.com --name Expert --role expert
$ specloop inbox
$ specloop expert:claim <request-id>
$ specloop answer <request-id> --comment "Add expiry checks and missing-token tests."
$ specloop get <request-id>