Technical founders & first engineers
You need to pick Pydantic AI vs LangGraph vs a vendor SDK this sprint, and nobody has time to read 80 READMEs and changelogs.
Custom technical research · AI agents · MCP · local-first infra
Your AI-stack decision, researched and sourced, in days, not weeks.
Tell us the decision: which agent framework, which MCP servers are safe to run, whether to self-host inference. You get a report built from primary sources, with every claim dated and linked and every unknown labelled as unknown.
Briefs from $300 · 3 business days · agent-researched, human-reviewed
Who it's for
Eng leads rolling out MCP
Security wants to know which MCP servers are maintained, how they're licensed, and which have known CVEs, before anyone connects them to real tokens.
Agencies & consultancies
A client wants a tooling landscape next week and your team is billable elsewhere. White-label is available on Landscape tiers.
Local-first & privacy-driven builders
Should you self-host inference, and on which engine? Ollama, llama.cpp, MLX, vLLM, compared against your hardware and constraints.
Technical diligence
Facts on an agent-infra company's stack vs. the open-source alternatives: maintenance, licences, momentum. Technical facts only, never investment advice.
Not a fit
Market-size / TAM estimates, security pentests, legal opinions, or anyone who needs a big-analyst logo for procurement.
Tiers & prices
Fixed prices. The clock starts once your intake is complete and payment is received.
Decision Brief
$300 one-time
One sharp question: "A vs B for our constraints?"
- ~1,500–2,500 words
- 15–30 primary, dated sources
- Executive answer + comparison table
- Risks & "could not verify" section
- Markdown, HTML & PDF
- 3 business days · 1 clarification round
Most complete per dollar
Landscape Report
$900 one-time
A category or stack decision, e.g. "our MCP stack".
- ~4,000–8,000 words
- 40–80 primary, dated sources
- Adopt / watch / avoid tables
- Licence, release & archive snapshot per option
- Security notes with CVE links
- 5 business days · 1 revision round
Landscape + Lab
$1,500 one-time
For a decision you want tested, not just read about.
- Everything in Landscape
- Top 2–3 picks installed & run in a clean Linux container
- Reproducible scripts + raw logs
- 30-minute call
- 30-day "what changed" addendum
- 7–10 business days
Add-ons: rush (≈half the turnaround) +50% · 30-day refresh $150 · extra revision $100 · white-label +$200 (Landscape tiers). Prices in USD.
Read a real report before you pay
This is the full MCP Server & Tooling Landscape (September 2026), a real deliverable with the client's identifying details removed. It covers the 2026-07-28 spec changes, SDK tiers, local-model hosts, servers by category, gateways, and a security section with real CVEs and mitigations.
It's a dated snapshot (2026-09-25). That's the point: every claim tells you when it was true and where it came from.
Open the sample report PDF (35 pages)“Unmaintained or archived hosts and servers: Continue (repo read-only), Roo Code (archived), mcphost (archived)… Don't install random npm MCP packages without pinning and scanning them; the malicious
postmark-mcpis the canonical example.”
How it works
1 · Intake
Answer a few short questions: the decision, your constraints (local-only? licence policy? budget?), and what you already ruled out.
2 · Scope & pay
You get a one-paragraph scope confirmation and a payment link. The clock starts when both are done.
3 · Research
A research agent reads primary sources (specs, docs, release feeds, NVD/OSV) and drafts. Nothing is cited that wasn't actually read.
4 · Human review
The recommendation is checked against your constraints and signed off by a person before it ships.
5 · Delivery
Markdown, HTML and PDF by email, then one clarification or revision round. The report is yours, and it's never published without written permission.
FAQ
Why not just use ChatGPT / Gemini / Perplexity deep research?
For a quick overview, you should. Those tools are cheap and fast. You pay for this when you'd otherwise spend a day verifying: we scope the question to your constraints, only cite pages actually read, check repos for archived flags, licence changes and last release dates, and list what we couldn't verify. Compare the sample with what your tool produces on the same question, and decide from that.
Who is behind this? Do you have references?
It's run by Remy (Jeremy Morgan-Jones Sr.) of LoveLogicAI LLC, an independent builder of local-first agent systems. This is a new service, so there are no client testimonials yet, and we won't invent any. What you can judge is the work itself: the full sample report is public on this page.
Won't this be out of date in a month?
Parts of it, yes. This ecosystem moves weekly. Every claim is dated so you know exactly what to re-check, each report has a "watch list" of things likely to change, and you can add a 30-day refresh addendum for $150.
Is AI used to produce the report?
Yes, openly. Research and drafting are done by an AI research agent working from primary sources. Scoping, the recommendation, and final sign-off are human.
What if the report has an error?
If a factual claim is wrong or unsourced, tell us and we'll correct it within 2 business days, free. If we can't deliver the agreed scope, you get a refund. See the refund policy sent with your scope confirmation.
Is this legal, security, or investment advice?
No. Licence notes are factual summaries, not legal opinions. Security notes cover public CVEs and advisories and aren't an audit of your systems. Nothing here is financial or investment advice.
Is my question confidential?
Yes. Your questions and report are never published or reused as a sample without your written permission.
Start a report
Send the intake. You'll get a scope confirmation and a payment link within 1 business day. Nothing is charged until you approve the scope.