Pitch decks, customer briefs, analyst reports, internal docs.
Sophyx Agent Memory
Train Sophyx like a five-year employee.
In 5 minutes.
Drop in PDFs, notes, and URLs. Sophyx compiles them into a living memory.md your agents read before every output.
/ ingesting
Live/ aggregated memory
memory.md
The personalization gap
Your agents are smart.
They just do not know you.
Without memory, your AI agents write content that could be about anyone. With memory, they write content that could only be about you.
Feed it everything you know
Five source types. One brain.
Whatever shape your brand knowledge is in, drop it in. Sophyx parses, stores, and rolls every source up into a single living memory.
- .pdf
PDF
- .txt / .md
Text
Brand voice guides, style rules, FAQs, positioning notes.
- .json
JSON
Pricing tables, product schemas, structured exports.
- .csv
CSV
Keyword sets, customer lists, content audits, ICPs.
- https://
URL
LinkedIn profiles, X threads, Reddit posts, any public page.
Every source is parsed, embedded, and saved in your workspace. Nothing leaves it.
How memory works
Three stages. From raw files to personalized output.
Sophyx Agent Memory is the same engine that powers your AI Gateway, your AI Marketing Agent, and every other Sophyx agent in your workspace.
- 01
Drop in sources
Upload PDFs and notes, paste JSON or CSV, drop a LinkedIn URL. Five accepted source types, no transformations required.
- 02
Sophyx builds the brain
Sources are parsed, embedded, and rolled up into a Knowledge Graph, a How-Internet-Sees-You analysis, and a ≤5000-character memory.md.
- 03
Every agent reads it
Sophyx agents load memory.md into context before producing anything. The output sounds like you, every time, on every channel.
Context · KG
Knowledge Graph
Entities, relationships, products, customers, differentiators. The structured graph every agent queries.
Context · Analysis
How Internet Sees You
A live analysis of how AI engines, search, and social currently describe your brand. Updated as new memories arrive.
Read by every agent
memory.md
The aggregated brand brief, capped at 5000 characters. Loaded into context before every output. Auto-refreshes on change.
Q: What is Sophyx Agent Memory?
A: Sophyx Agent Memory is the hyper-personalization layer for Sophyx AI agents. A workspace owner drops in PDFs, notes, JSON, CSV, or URLs. Sophyx compiles them into a Knowledge Graph, a How-Internet-Sees-You analysis, and a ≤5000-character memory.md. Every Sophyx agent reads that file before generating content, so the output reflects the brand instead of generic LLM averages. The AI Gateway consumes the same memory, end to end.
Same prompt. Different brain.
Before memory. After memory.
One agent. One brief: write a LinkedIn post about our launch. The only thing that changes is whether the agent has memory.
Without memory
“Excited to announce our new product! We are leveraging cutting-edge AI to deliver innovative solutions that empower businesses. Stay tuned for more updates as we continue our journey. #AI #Innovation”
- Could be about literally any company.
- No proof. No specifics. No voice.
- Reads like a press release written by a robot.
With Sophyx memory
“5,200 brands later, here is what we know: a Series B SaaS that shows up in the ChatGPT answer closes 28% more pipeline than one ranking #1 in Google. So we built the AI Gateway: a machine-readable replica of your site at llms.yourbrand.com, live in 14 days. Demo this week, link in comments.”
- Pulls in real proof (“5,200 brands”) from memory.
- Matches your voice: direct, technical, zero hype.
- Names the actual product. Names the actual outcome.
What you actually get
One workspace. Every agent personalized.
Agent Memory is included in every Sophyx plan. It is the upstream context layer that powers your AI Gateway and every agent that produces content for your brand.
- The flagship
memory.md, the brand brief every agent reads
A single ≤5000-character Markdown file aggregated from all your memories. Loaded into the context window of every Sophyx agent on every run. The single source of truth.
Auto-refresh on change
Add or update a memory, and memory.md rebuilds. Every downstream agent immediately uses the new context. No re-prompting.
5 source types
PDFs, text, JSON, CSV, and URLs (LinkedIn, X, Reddit, any public page). One workspace, every shape of brand knowledge.
Live Knowledge Graph
Entities, relationships, and facts about your brand, structured for retrieval. Used by content generation and your AI Gateway.
How-Internet-Sees-You analysis
A live read on how AI engines, search, and social currently describe you. Surfaces gaps memory needs to close.
Powers every Sophyx agent
The AI Marketing Agent, the Content Generation pipeline, the Prompt Optimization tester. All of them read memory.md first.
Feeds the AI Gateway
The same memory powers the machine-readable replica of your site at llms.yourbrand.com. Brand personalization stays consistent across AI agents and AI engines.
Per-workspace isolation
Your memories live only inside your workspace. Never used to train third-party models. SOC-grade access controls and audit trail.
Frequently asked
What buyers ask before training memory
Stop sounding like every other brand
Train your agents. In 5 minutes.
Drop in your first three sources. Sophyx builds the memory. Every agent in your workspace starts sounding like you on the very next run.
- 5 source types accepted
- Auto-refresh on change
- Read by every agent
See related: Sophyx vs Peec AI · Sophyx vs Profound · Sophyx vs SearchAtlas · AI Gateway · AI Marketing Agent · Content Generation · Prompt Optimization · AEO & GEO · AI Visibility Tracker