Thread Transfer
Introducing Thread Transfer: Smarter context management
A look at the prototypes that failed, what finally resonated with operations teams, and how we're building a durable context layer for AI.
Jorgo Bardho
Founder, Thread Transfer
Thread-Transfer exists because teams kept coming to us with the same complaint: “We finally convinced our org to use AI in production, and now all of our context lives in one tool that nobody else can access.” When conversations drive roadmap decisions, support escalations, audits, and customer renewals, locking that context into a vendor-specific transcript is a non-starter. We needed a portable, verifiable way to move threads across systems without watering them down. This post explains what we built, the painful lessons that shaped it, and where we’re taking the platform next.
Where the idea came from
Our founding crew spent a decade building developer tooling and enterprise messaging products. Every rollout ended with the same awkward spreadsheet: rows of conversation exports, coloured cells marking who copied what into a doc, and a column labelled “did we lose the decision?” Meanwhile, finance teams tried to reconcile token invoices that had no traceability back to a specific project. Watching brilliant teams drown in copy-and-paste convinced us the internet needed a context layer that respected the original conversation but travelled anywhere.
Prototypes that failed
We built three prototypes before bundles clicked:
- The raw export. We streamed chat logs into a JSON file with minimal structure. Customers shrugged. It was easier to scroll Slack.
- The AI summary bot. We summarised every conversation and emailed a recap. Without references or provenance, nobody trusted it.
- The notebook view. We split conversations into headings inside an internal notebook app. Great for demos, useless for production workflows.
We scrapped them all and asked a sharper question: what would it take for a compliance officer, an engineer, and a customer success manager to call the same export “the source of truth”?
The product principles that survived
Four non-negotiables shape Thread-Transfer bundles:
- Deterministic outputs. If you feed the same thread through the pipeline tomorrow, you get an identical bundle. Every transformation step is logged and exposed.
- Rich structure with plain JSON. Bundles are deeply structured—blocks, references, manifests, hashes—but still readable in any text editor.
- Verifiable provenance. Every block links back to the original message ID, timestamp, and author.
- Inline metadata for automation. The same file should drive audit trails, knowledge bases, and LLM prompts without custom parsing.
Under the hood
The core pipeline is orchestrated by our distillation service. It ingests raw transcripts from chat, call-centre tools, or exported CSVs and runs several passes:
- Signal extraction. Key claims, decisions, blockers, and follow-ups surface via heuristic + model-assisted ranking.
- Deduplication and normalisation. Redundant statements collapse, acronyms expand, and numeric references align to existing systems.
- Schema packaging. Content slots into the bundle schema—facts, decisions, narratives, references, metadata—while maintaining links to source messages.
- Integrity sealing. Outputs ship with a SHA-256 hash and manifest so teams validate the bundle before ingesting it.
Real-world deployments
Here are three ways customers use Thread-Transfer today:
- Incident retrospectives. Engineering leads export a Slack or Teams incident channel into a bundle, archive it, and feed it into their retro workflow. Deterministic outputs mean they can diff versions when new facts emerge.
- Customer escalations. Success managers attach bundles to Salesforce cases so legal, finance, and product all read the same narrative without pinging frontline agents.
- Knowledge base upkeep. Product marketers pipe bundles into Notion as the first draft of a changelog or release note, complete with verified quotes and decisions.
Results so far
Across the first cohort of customers, bundling cut context windows by an average of 57%. Teams report 30–40% faster hand-offs between support and engineering, and finance finally gets the token-level data it needs to forecast invoices. Most importantly, compliance teams sign off on bundles as an audit-ready artifact.
What’s next
- Layered access controls so sensitive blocks redacted automatically.
- Diffable bundles for recurring conversations (weekly reviews, long-running deals).
- Deeper integrations with knowledge graph and analytics tooling.
Thread-Transfer will keep evolving, but the mission stays the same: make it effortless to move conversations to the tool that needs them without losing fidelity, accountability, or time.
Join the waitlist
We’re onboarding teams in waves to keep quality high. If you want in on the next batch, join the waitlist or email me at info@thread-transfer.com. Bring your messiest threads—we’ll help you turn them into durable context.
Learn more: How it works · Why bundles beat raw thread history