La plateforme de connaissance et mémoire pour les agents IA de code.

Recherche sémantique, mémoire persistante, règles d'organisation et synchronisation de dépôt dans une plateforme native MCP.

Precise context. Not a wall of text.

Indexed Knowledge

Elastra ingests your repositories and documentation, chunks and embeds them, and serves them back as precise, filtered semantic search results. Agents get exactly what's relevant — nothing more.

→Automatic chunking and embedding on every push
→Hash-based deduplication — no reindexing unchanged content
→Freshness signals — results prioritize recently updated content
→Filtered by organization, project, and namespace

knowledge

# Agent queries codebase knowledge
elastra_context(query: "how does auth work")
→ Returns top 5 relevant code chunks
 
# Result includes freshness info
{ "score": 0.94, "path": "internal/auth/service.go",
  "freshness": "indexed 2h ago" }

Decisions that survive every session.

Shared Memory Layer

Agents can write memories — decisions, conventions, debugging conclusions — that persist and are shared across the team. The next run starts where the last one left off.

→Memory scoped to organization and project
→Automatically surfaced alongside knowledge search
→Agents in a pipeline share a single truth
→Audit trail: know what was remembered and when

memory

# Agent writes a memory after debugging
elastra_memory_write(
  type: "bug",
  title: "Redis TTL causes stale sessions",
  content: "Session cache TTL is 5min but refresh is 6min.
            Fixed in PR #482."
)
 
# Next agent session retrieves it automatically
→ "Known issue: Redis TTL mismatch (see PR #482)"

Always up-to-date. Zero manual steps.

Repository Auto-Sync

Connect GitHub or GitLab repos from the Elastra console. Webhooks keep the index fresh on every push. Status is always visible — no guessing.

→GitHub and GitLab support out of the box
→Auto-sync toggle per repository
→Index status: queued → processing → indexed
→Webhook diagnostics in the console

sync

# Connect via CLI
elastra auth login
elastra init
# → Links workspace to Elastra project
 
# Or via console: Repositories → Connect
# After each git push:
→ Webhook received
→ Changed files queued for indexing
→ Status: indexed ✓

Every agent follows your standards.

Org-Level Rules & Personas

Define coding conventions, style guides, and agent behavior at the organization or namespace level. Rules are injected into every agent response automatically.

→Organization-wide rules applied to every agent
→Namespace-level overrides for specific projects
→Agent personas for consistent behavior
→Versioned and auditable

rules

# ELASTRA.md — project contract
## Rules
- Always use typed errors in Go
- Follow REST conventions for new endpoints
- Use structured logging (zerolog)
 
# Agents receive these rules automatically
→ elastra_rules() → injects org + namespace rules

Full control at every level.

Console & CLI

The Elastra console manages your workspace. The CLI connects your local machine and configures AI agents. Both are part of the same seamless experience.

→Console: manage projects, repos, usage, roles
→CLI: install, authenticate, connect, diagnose
→19+ agents supported via `elastra init`
→Works on macOS, Linux, Windows

cli

# CLI commands
elastra auth login        # Authenticate your account
elastra init             # Configure all agents
elastra doctor           # Verify everything works
 
# Console: app.elastra.ai
→ Projects → Repositories → Chat → Settings

Know exactly what your agents did.

Audit & Observability

Every agent action is logged with full context. Audit logs are available in the console. For Enterprise, structured logs integrate with your observability stack.

→Audit log for every agent action
→Query by agent, project, or time range
→SCU usage tracking with real-time alerts
→Enterprise: structured JSON logs via Loki/Grafana

observability

# Usage alert example
GET /v1/usage
→ {
    "scu_used": 42000,
    "scu_limit": 50000,
    "warnings": [{
      "type": "scu_warning",
      "message": "84% of monthly SCU used"
    }]
  }

Voir en action

Commencez gratuitement et connectez votre premier dépôt en quelques minutes.

Commencer gratuitement Lire la documentation