The short version
Three steps, every time.
1
Fingerprint
Deterministic stack extraction
Blazer reads manifests, lockfiles, Dockerfiles, and infra configs to build a capability fingerprint. Fingerprints get a unique ID from a one-way hash based on your repo and calculated client-side. No source code leaves your machine.
HMAC-SHA-256 · fp_7a3c2e9b4f81
2
Match
Archetype-weighted retrieval
Application fingerprint is matched against dozens of tech stacks and architecture patterns. Products in the catalog are scored against your applications's stack, not keyword overlap or stale training.
fp_7a3c2e9b4f81 · NextJS · Postgres · Clerk · Terraform IaC w/ AWS
3
Reason
Citable decision receipts
Every recommendation ships with structured metadata — a compatibility score, integration-quality signals, and provisioning and pricing context. Every call lands in the journey log. Your architecture reviews read the receipt, not a chat transcript.
Journey log · Structured review data
1
Fingerprint
Reads your repo
Parses your package.json, Gemfile, pyproject, Dockerfile, terraform — whatever's there. Builds a capability map: "Next.js + Postgres + Redis, B2B SaaS shape." Takes a few seconds, runs locally. No source code leaves your box.
Local · HMAC-SHA-256
2
Match
Picks the right shelf
Compares your stack to dozens of archetype patterns and ranks products whose capability tags line up with the shape you're actually running — not the ones with the best SEO.
200k+ products · dozens of archetypes
3
Reason
Explains itself
When your agent suggests a product, you get a compatibility score, integration-quality signals, and a journey log your team can replay. Beats "trust me, it's good."
Structured JSON · journey-logged
1
Look
Blazer looks at your project
When you ask Claude Code to add something, Blazer takes a quick look at your project to figure out what you're building — like a friend glancing at your screen to get the vibe before giving advice. Your code stays on your machine.
Takes about a second
2
Match
Finds stuff that fits
Blazer has a huge list of tools and services that other developers use. It picks the ones that fit projects like yours — not the ones that sound fanciest.
Way more tools than you could try yourself
3
Explain
Tells you why
When Claude suggests a tool, you get a plain-English explanation: how well it fits your project, what it's good at, what it'll cost. No more "I have no idea why it picked that."
Readable, not magic
1
Identify
Know what your teams are building
Blazer fingerprints each codebase and maps it to a stack archetype. A tenant-scoped view of who's on what — this many Next.js B2B apps, that many Python data platforms, and so on — without reading a line of code.
Tenant-scoped rollup
2
Align
Shape what your agents pick
Agents pull from a shared catalog that lives in your tenant alongside the public one. Over time, what your teams actually choose — and what they bypass — becomes visible in the journey log, so architecture standards evolve from data rather than opinion.
Org-scoped tenant
3
Audit
Every agent pick is reviewable
Every blazer.find() call lands a structured JSON record in the journey log: what the agent asked, what Blazer returned, what was picked. Your architecture board replays it; compliance exports it. Nothing lives only in a chat transcript.
Structured journey logs
Reasoning receipt
Every pick comes with its homework.
Every blazer.find() returns a compatibility score, integration quality signal, and journey context — structured JSON your team can log, not a black-box number.
Every blazer.find() returns a compatibility score, integration-quality signal, and journey context — structured JSON you can log, diff in a PR, or argue with. Same shape your teammate sees.
When Claude Code suggests a tool, you see the category it fits, how well it matches your project, and what it's good at — not just a library name you've never heard of.
Every blazer.find() call lands a structured JSON record in the journey log — what the agent asked, what Blazer returned, what was picked. Replay it in review; audit it later.
Archetype affinity0.94
Capability overlap11 / 12
Reviewer similarity8 matches
Final score91%
The concepts
The five ideas behind Blazer.
fingerprint
Codebase fingerprinting
A deterministic capability graph extracted from your repo — runtimes, frameworks, data stores, auth patterns, deployment target. Stored as a content-addressed hash so two agents on the same commit see the same registry view.
archetypes
Archetypes
Dozens of canonical stack shapes — "B2B SaaS on Next.js + Postgres + Clerk", "Data platform on Python + Snowflake + Airflow", etc. Every product in the catalog declares which archetypes it fits, weighted by production-grade evidence.
catalog
Product catalog
200,000+ products indexed and growing. Each is categorized and tagged with machine-readable capabilities your agent can reason over. Coverage depth varies — widely-used products carry richer structured facts; long-tail entries still get you a ranked shortlist.
reviews
Reviews & ratings
Reviews are weighted by stack-archetype similarity to your own. A Postgres-shop review of Stripe counts for more in a Postgres archetype than a Firebase review does. Reviewer identity is tied to a GitHub-verified fingerprint.
matching
Capability matching
Each result carries a compatibility score against your stack alongside structured signals — integration quality, provisioning, pricing, and recorded journey context. Enough for your team to audit the decision rather than take a single number on faith.
fingerprint
Codebase fingerprinting
A read-only scan that produces a capability map of your repo — runtime, framework, storage, auth, deployment target. Runs in a few seconds. No source code leaves your box; only the capability signals and hashed identifiers go over the wire.
archetypes
Archetypes
Pre-classified stack patterns like "Next.js B2B SaaS" or "Python data platform". Products declare which archetypes they fit; your fingerprint is matched against the archetype that looks most like you.
catalog
Product catalog
Hundreds of thousands of SaaS products sorted by what they do and how they plug into a stack. Structured where it matters — capabilities, category, integration signals — so your agent picks on merit, not marketing.
reviews
Reviews & ratings
Reviews come with structured context: the reviewer's stack, archetype, what they actually used the product for. Weighted toward reviewers whose setup looks like yours — the goal is relevance, not volume.
matching
Capability matching
Every result ships with a compatibility score against your stack plus structured signals — integration quality, provisioning, pricing, journey context. Your agent sees more than a single number; you can diff the signals when a pick looks weird.
fingerprint
Fingerprinting your project
It's like a quick photo of what your project looks like — React? Python? Got a database? Blazer uses the photo to give your agent context. No code and no secrets leave your machine.
archetypes
Project types ("archetypes")
There are patterns in software. A SaaS app looks different from a mobile game. Blazer knows dozens of common patterns and figures out which one your project looks most like, so its suggestions actually fit.
catalog
The catalog
A huge, sorted list of software — payments, auth, databases, email, you name it. Organized so your agent can tell what each one actually does, not just that it exists somewhere on the internet.
reviews
Reviews from real people
Reviews come with context — what the reviewer was building and what stack they ran. When Claude looks at reviews, it pays more attention to the ones from projects like yours, so the advice feels relevant instead of generic.
matching
Finding the right one
When Claude asks Blazer for tools, each answer comes with a match score for your project — so you can tell which ones actually fit, not just which ones showed up.
fingerprint
Codebase fingerprinting
A deterministic signature of each codebase's technical shape. Lets you roll up — "60% of new repos this quarter are Next.js B2B" — without reading a single line of code.
archetypes
Archetypes
Dozens of canonical stack patterns. Teams sharing an archetype share a vendor shortlist. Standardization without mandating a monoculture.
catalog
Product catalog
200,000+ products indexed and growing, categorized by purpose and tagged with machine-readable capabilities. A searchable substrate for vendor decisions — your teams pull from the same catalog your architecture team reviews.
reviews
Reviews & ratings
Reviews carry structured context — reviewer stack, archetype, what the product was used for. Weighted toward matches on your stack, so a fintech review of a payment provider counts differently than a hobbyist's does.
matching
Capability matching
Each result ships with a compatibility score and the structured signals beside it — integration quality, provisioning, pricing — plus a journey-log entry. Compliance can trace any decision back to the signals that accompanied it.
FAQ