Research, case law, contracts, filings: legal agents that read your exhibits, draft, and cite every source — checkable line by line.
From research to drafting, from the file to compliance — specialised agents that do the work, grounded in your exhibits.
Agents per practice area — contracts, litigation, employment, real estate, family, compliance — that qualify, research and draft.
Answer from your own exhibits and codes, every claim cited and checkable.
Formal notices, pleadings, contracts, letters — in the firm’s format, ready to review.
Drop the exhibits: timeline of facts, parties, missing documents and the points of law.
Track the texts that concern you and keep the in-force version straight.
Built for a practice: partitioning, human approval, and deployment inside your walls.
No waiting list and no hidden paid module: this is the full set of capabilities available today.
Contracts, litigation, employment tribunals, leases, family, consumer, company and tax — one click each.
Build an agent around your speciality, with its instructions, tools and your knowledge bases. No code.
Your matters, codes and templates become a queryable base, with cited sources.
The Algerian legislative corpus, queryable: article-level grounding, amendment tracing, in-force version.
An orchestrator splits the matter across several agents and shows you every step live.
Documents, letters and contracts generated in your house style, editable then exported to DOCX or PDF.
Live search over the sources you choose, with verification that the version is still in force.
Hearing and meeting transcription, translation of whole exhibits, multilingual dubbing.
Periodic monitoring, deadline reminders and reports emailed to you, with no intervention.
Sensitive actions (sending a letter, publishing) wait for your explicit approval.
Microsoft 365, Google, Notion, and your own practice tools through API keys.
Publish an agent as a chatbot on the firm’s website, in your branding, in minutes.
Team accounts, shared matters and bases, roles and seats managed from the interface.
Per-account partitioning, sandboxed execution, audit log, GDPR export and deletion.
Personal API keys, durable runs that continue with the tab closed, and usage tracking.
Interface and agents in French, Arabic, English and Chinese — Arabic is first-class.
Vector + full-text + entities + graph, on a 100% Rust stack. Article-level grounding, amendment tracing, every answer cited and checkable.
p50 167 ms at 50 concurrent · 3× faster than a Python stack · replayable benchmarks.
Any single retrieval method misses something. All four run in parallel and are fused deterministically: same question, same answer.
→ RRF fusion + entity boost + MMR diversity → ranked, sourced answer
Clickable [N] citations that open the exact PDF page or the highlighted section.
It scores each claim against its sources and says "insufficient evidence" rather than inventing one.
Follows a fact across versions — created, amended, repealed — and separates "in force" from "back then".
Ask for a product by description and get the photo itself, linked to the page describing it.
100 MB+ / 1,500-page catalogues via batch OCR. PDF, DOCX, PPTX, XLSX, images.
RTL, broken plurals, hamza/tashkeel normalisation — the same accuracy as English.
Tested, measured, replayable
Finance suite 12/12 · legal codes 10/10 · +11 points against EdgeQuake on MMLongBench. Domain-tuned profiles (legal, corporate, timeline, catalogue) and deployable on-premise or air-gapped.
A general assistant talks. Legal-Index works the file — grounded, sourced, checkable, and deployable inside your practice.
Answers grounded in YOUR documents
✕ Ungrounded, can hallucinate
✓ Ultra-RAG — every claim cited
Expert agents per domain
✕ One general assistant
✓ 50+ specialized agents
Takes real actions
✕ Chat only
✓ Tools + human-in-the-loop approval
Your data isolation
✕ Shared & opaque
✓ Private, encrypted, sandboxed
Team & company controls
✕ Personal use only
✓ Orgs, seats, sharing, SSO
On your own servers
✕ Not possible
✓ On-premise fully available
Languages
✕ Limited RTL
✓ EN · FR · AR · ZH, full RTL
Real chains: qualification, research, drafting, verification — through to the delivered document.
The augmented practice
Describe the matter: the agent qualifies it, researches the grounds, builds the strategy, drafts the document, then a verifier checks every citation before handing it back.
The pipeline
‖ in parallel — one agent per angle
↺ returns the work while a citation is missing
The investigation team
Drop in every exhibit: the team reads them, reconstructs the timeline of facts, identifies the parties and points of law, flags what is missing and returns a structured note.
The pipeline
‖ in parallel — one agent per angle
↺ flags the missing exhibits
Continuous watch
Pick your practice areas: the watch follows published texts, filters the noise, checks whether the version is still in force and sends you a dated note.
The pipeline
‖ in parallel — one agent per angle
↺ filters the noise before publishing
This is only a glimpse.
See the full catalogue →The catalogue is a starting point. Compose an agent around your speciality, then take it out: onto the firm website, into your tools, or entirely onto your own infrastructure.
The Agentic Studio turns an agent into data: a name, instructions, and capabilities you flip on like switches. Each capability brings its tools AND the know-how that guides how they are used.
An agent becomes a public chatbot on your own domain. Paste a script, pick the allowed domains, and it's live — your visitors' conversations bill to your account, not theirs.
<script
src="https://ai.syctra.com/embed.js"
data-agent="support-fr"
defer
></script>Create an API key in your settings, exchange it for a short-lived token, and call any endpoint — the same ones the interface uses. Your agents become a component of your own product.
curl https://ai.syctra.com/api/v1/auth/token \
-H "X-API-Key: sk_live_…"
curl https://ai.syctra.com/api/v1/chat \
-H "Authorization: Bearer $TOKEN" \
-d '{"agent":"my_analyst",
"message":"Résume le rapport Q3"}'Four real matters, from file to document — with the agents involved at every step.
The pipeline
The pipeline
The pipeline
The pipeline
Three steps, from the ask to the delivered file.
Two sentences are enough. Pick the team, attach your documents or a knowledge base, and start.
Each agent reasons live in the thread. The verifier sends the work back until the result is good enough.
Deck, report, spreadsheet or site: downloadable immediately and kept in your history.
Start free. Upgrade as you grow. Credits are spent only on what you use.
How credits work
1,000,000 tokens = 1,000 credits
You only pay for what you use — no wasted credits.
＋ Need more? Top up your credit anytime — any custom amount.
What people ask before getting started.
A credit is the platform’s unit of work. A mission consumes credits based on the effort the agents actually spend — research, writing, verification loops. A full presentation typically costs a few dozen credits.
No. Purchased credits stay yours and roll over month to month. You only pay for what you use.
Yes. Every account is partitioned: documents, knowledge bases, conversations and logs. Agent code runs in a sandbox, and sensitive actions go through human approval.
Yes, completely. The full stack installs on your machines, including on a network cut off from the internet, with your own models (vLLM, Ollama) or your API keys. No data leaves your perimeter.
Yes. Managed models (Gemini, OpenAI, Anthropic, Mistral) or self-hosted ones (vLLM, Ollama, Hugging Face, LM Studio): any OpenAI-compatible endpoint works, and the choice is made model by model.
Yes, with no commitment and no fee. Credits you already bought stay usable after cancelling.
What we ship, version after version. Every line is already in production in firms.
Create your account in seconds — or let’s talk about a deployment on your own servers.
SYCTRA is not limited to public SaaS: it deploys on your systems, under your brand, on the AI engine you choose.
The platform runs on your servers or private cloud. No data leaves your perimeter — decisive in health, defence, legal and the public sector.
Your logo, your domain, your brand. We provide the engine, you sell the platform to your clients under your own name.
Agents built for your trade: we write their instructions, tools and quality checks together with your experts.
Connect the platform to your systems: API keys, run queues, SSO, audit logs and export to the tools you already use.
In air-gapped mode nothing crosses this boundary: no document, no query, no log.
▸ One Docker stack, one command
A complete, versioned, documented Docker Compose stack. Install, upgrade and restore follow the same procedure we run in our own production.
▸ Your models, your keys
Point it at vLLM or Ollama to stay fully local, or use your own API keys for hosted models. The choice is made model by model.
▸ Air-gap, for real
No outbound call is required: no telemetry, no online licence check, no third-party service needed to boot.
▸ You own everything
Documents, vector indexes, conversation history and audit logs live in your database, on your disks, under your backups.
Managed APIs or self-hosted models — any OpenAI-compatible endpoint works.
A specific need, a large volume, or a reseller project?
Let’s talk about your project →A question, a demo, a partnership, an investment — or anything else on your mind? Tell us, don't hesitate. We read every message and reply fast.
Custom teams
Want a team of agents dedicated to your trade? We build it with you, from defining the agents to deployment.