DeepJudge + Legora Form ‘Bidirectional Partnership’ – Updated

DeepJudge and Legora's "Bidirectional" Deal Is Really About Whose Server Your Institutional Knowledge Lives On

DeepJudge, a knowledge-management vendor built on retrieval-augmented search, and contract AI platform Legora have announced what both companies call a "bidirectional partnership," first reported by Artificial Lawyer on 18 August 2026. The deal builds on DeepJudge's Agent Handoff Protocol (AHP), which Thomson Reuters and Harvey had already agreed to adopt. Under the new arrangement, a lawyer can pull precedent and prior work surfaced by DeepJudge directly into a Legora drafting or negotiation session, and — the "bidirectional" part — new work created inside Legora flows back into DeepJudge so it becomes searchable institutional knowledge for the next matter. DeepJudge co-founder and CEO Paulina Grnarova and Legora co-founder and CEO Max Junestrand both put out statements framing this as knowledge "compounding" over time rather than sitting static in a repository.

What actually changes for a lawyer using these tools?

Search and drafting stop being separate steps — a lawyer finds precedent in DeepJudge and sends it to Legora, or Legora pulls it automatically, without leaving the workflow.

Concretely: a lawyer working in Legora on a negotiation or review task can draw on precedent, prior deals and firm experience that DeepJudge has indexed, instead of separately searching a DMS or asking a colleague. Once that work product exists, DeepJudge can index it too — along with anything else stored in Legora projects — so it shows up in future searches across the firm's other knowledge sources. That's the "one search across everything" DeepJudge describes.

Is this different from the Agent Handoff Protocol announced the week before?

Not really — AHP is the plumbing, this partnership is the first fully-built example of two vendors' products actually using it together.

DeepJudge's own spokesperson initially implied the Legora tie-up went further than what Thomson Reuters and Harvey had signed up for via AHP, then walked that back, clarifying it was "very much connected" to the AHP story but represented "the next chapter" — a deeper product integration rather than a separate protocol. Worth noting the correction happened the same day as the announcement, which is a small tell about how firm the messaging was to begin with.

What should litigation and e-discovery teams actually ask before signing on?

Ask exactly what "institutional intelligence" DeepJudge is indexing from Legora projects, who can see it, and whether any of it touches privileged matter data.

Neither company's statements say whether documents flowing between the two platforms include privileged litigation materials, how access permissions carry across the handoff, or whether indexed "work product" is scoped per-client, per-matter, or firm-wide by default. For a KM tool that's a reasonable question to leave open; for anything touching discovery material it isn't. If your firm handles both transactional and contentious work on shared infrastructure, get a straight answer on tenancy and training-data boundaries before any evidence-adjacent content crosses that bridge — questions this announcement doesn't address at all.

Frequently asked questions

Does this partnership involve e-discovery or evidence data directly?

No — both companies describe this as a knowledge-management and drafting integration for transactional and advisory work, not a litigation evidence platform.

Do other vendors use the same protocol?

Thomson Reuters and Harvey have both agreed to adopt DeepJudge's Agent Handoff Protocol, according to DeepJudge; Legora is described as the first to build a deeper product integration on top of it.

Is any performance or adoption data available yet?

No figures were published in the announcement — only architectural and workflow claims from the two CEOs, with no independent measurement cited.

Sources: Artificial Lawyer, "DeepJudge + Legora Form 'Bidirectional Partnership' – Updated," 18 August 2026.