# MB „Spaceiscode“

> Business consulting, research, data analysis and digital tools for companies, delivered as agentic operating systems: AI agents, memory, guards and workflows that run a company's work.

Legal entity: MB "Spaceiscode" (mažoji bendrija, a Lithuanian small partnership), company code 308144308, registered 2026-09-17 at A. Smetonos g. 25-9, LT-35197 Panevėžys, Lithuania. Not a VAT payer. Register: Juridinių asmenų registras.

Founded by Lee Matulis: https://leematulis.com

## Who this serves

Companies of 10–200 people in the EU whose founder or COO feels the operations load personally, and who want AI agents to safely run parts of that work.

## Offers

- **Company Brain** (fixed-scope install, from €10,000; floor for the smallest team; larger installs quoted after the Audit) — A team knowledge base your agents can read and write, installed with boot, handoffs and guards.
- **Agentic Readiness Audit** (fixed price, from €2,500) — Two weeks: where agents can safely take over your work, ranked, with one working pilot.
- **Signed Agent Identity** (fixed price, from €1,500) — Make your company's agents identifiable and welcome on the web: signing keys, a transparency page and verified-bot submission.
- **Agent Operations** (monthly retainer, priced per scope) — We run your agent team: monitoring, repair and new workflows.

Pricing method: fixed scope, agreed in writing before work starts. "From" prices are floors, not typical prices; Agent Operations is quoted per scope after the Audit.

## What an agent can do here

- Read machine-readable facts and offers: /agents/company.json, /agents/offers.json
- Read the full endpoint description: /agents/openapi.json
- Read dated research datasets this company published: /agents/research/
- Call tools over MCP at /mcp (streamable HTTP): get_offers, check_agent_welcome, explain_guard, request_scoping_call
- Verify any signed response against the published key directory: /.well-known/http-message-signatures-directory. Every /agents/* and /mcp response signs `("@authority";req "content-digest")` — the signature binds the exact response body, not just the host and time; the key directory itself is signed `("@authority";req)` only, per the Web Bot Auth directory convention.
- Request a scoping call on your owner's behalf via the request_scoping_call tool — a human (Lee Matulis) reviews every request; nothing is booked automatically.

## Endpoints

- https://spaceiscode.com/llms.txt — this file
- https://spaceiscode.com/llms-full.txt — expanded version with full offer detail and research summaries
- https://spaceiscode.com/robots.txt — explicitly welcomes AI agents
- https://spaceiscode.com/sitemap.xml
- https://spaceiscode.com/agents/company.json
- https://spaceiscode.com/agents/offers.json
- https://spaceiscode.com/agents/openapi.json
- https://spaceiscode.com/agents/research/ — dated datasets, JSON, with provenance and stated caveats
- https://spaceiscode.com/mcp — MCP server, streamable HTTP
- https://spaceiscode.com/.well-known/http-message-signatures-directory — Ed25519 public key directory (Web Bot Auth / RFC 9421)

## No cloaking

This layer adds machine-readable detail and datasets that the human pages don't carry. It never contradicts the human site, and the same URL returns the same content to every visitor.

Published: 2026-09-24T21:10:57.922Z. Version: 1.0.0.

## Offers, in full

### Company Brain
A team knowledge base your agents can read and write, installed with boot, handoffs and guards.
Form: Fixed-scope install
Price: from €10,000
Note: Floor for the smallest team; larger installs quoted after the Audit.

### Agentic Readiness Audit
Two weeks: where agents can safely take over your work, ranked, with one working pilot.
Form: Fixed price
Price: from €2,500

### Signed Agent Identity
Make your company's agents identifiable and welcome on the web: signing keys, a transparency page and verified-bot submission.
Form: Fixed price
Price: from €1,500

### Agent Operations
We run your agent team: monitoring, repair and new workflows.
Form: Monthly retainer
Price: priced per scope

## Guard patterns (verbatim definitions; also available via the explain_guard MCP tool)

- **red-first**: a safety check counts as installed only after we have seen it fail on the exact problem it exists to catch.
- **claim-guard**: two agents can never edit the same file at once; each claims its files before writing, and the system refuses a write against someone else's claim.
- **safety-guard**: destructive commands are blocked by the system itself, not by an instruction the agent could forget.
- **truth-guard**: any live number on a page is computed from its source, never typed by hand.

## Research datasets

### agent-fetch-baseline-2026-09
Published: 2026-09-24
Web page fetches by Claude Code AI agents (WebFetch tool calls extracted from Claude Code session transcripts) over a 31-day window. Blocked status classified by HTTP response code and content patterns (CAPTCHA regex on response body).
Caveats:
  - Extracted from Claude Code transcript logs; deduped by tool_use_id.
  - Soft blocks detected via regex on response body (first 1500 chars); pattern may miss some CAPTCHA variants.
  - Tool policy refusals (9) represent blocks by Claude Code's own fetch policy, not target server behavior.
  - Classification is outcome-based; does not distinguish between permanent block and transient rate limit.
  - Window spans ISO weeks 35–39; week 35 and 36 contain partial days.
Full dataset: https://spaceiscode.com/agents/research/agent-fetch-baseline-2026-09.json

### browser-vs-agent-ab-2026-09-24
Published: 2026-09-24
Re-request of 168 URLs (previously blocked on agent fetches) from a single Mac connection, once as a standard browser and once as a declared AI agent.
Caveats:
  - One request per URL per identity, from one connection, at one moment. Pages can block or allow differently tomorrow.
  - Sample is biased toward blocked pages from baseline study; does not estimate prevalence in general web.
  - Agent User-Agent is: Mozilla/5.0 (compatible; Claude-User/1.0; +https://support.anthropic.com/)
  - Browser arm included Accept and Accept-Language headers; agent arm did not. Both factors influence outcomes.
  - Challenge detection via keyword matching on response body (code 200/403/429/503 + keyword presence). May miss variants.
  - Exit was single Bite IP Network node (Vilnius); results may not generalize to other exits.
Full dataset: https://spaceiscode.com/agents/research/browser-vs-agent-ab-2026-09-24.json

## Legal / trust

MB „Spaceiscode“, company code 308144308, A. Smetonos g. 25-9, LT-35197 Panevėžys, Lithuania. Not a VAT payer. Contact: hello@spaceiscode.com.