Project Intelligence model
Project Intelligence is Snipara's judgment layer for AI agent work. Before an agent edits, it should know what changed, why, what it impacts, what should happen next, and whether evidence says proceed, review, or stop.
Project Intelligence formula
Context
Source-backed project facts, code graph context, workflow state, and freshness metadata.
Decisions
Reviewed rationale, active constraints, confidence, authority status, and stale warnings.
Outcomes
Guard, review, test, deploy, and workflow evidence that can become typed local calibration receipts before any stronger outcome claim is made.
Judgment
Advisory next actions with confidence, evidence, counter-evidence, caveats, and calibration limits.
The product is intentionally broader than memory. Snipara gives humans, Claude Code, Codex, Cursor, CI, and custom agents one governed project layer to inspect before acting.
Product boundary
Snipara is not the reasoning model. Claude Code, Codex, Cursor, ChatGPT, and customer agents still reason and execute. Snipara supplies the project-owned structure: context, reviewed memory, source authority, impact, workflow state, coordination, and verification guidance.
Homepage relationship
The homepage leads with Project Intelligence. Workflow Continuity is the execution mechanism, outcome-weighted judgment is the proof, and memory, coordination, governance, collaboration, and verification are supporting inputs and surfaces. This page explains how Snipara answers the five agent-work questions while keeping emerging behavior and product limits visible.
The five questions
Memory alone is not enough for serious agent work. A useful project layer must answer these questions in a way that a human can inspect and a model can act on.
What changed?
What Changed For Me, Team Sync handoffs, PR Answer Packs, resume context, recent files, and workflow journals summarize repository movement before the next session starts.
Why?
Reviewed decisions, issue links, PR and MR Rationale Packs, handoff notes, and Why Capture keep source-authored rationale attached to project evidence and pending human review.
What does it impact?
Code graph context, impact plans, affected symbols, related tests, routes, and config facts give agents a blast-radius view before editing.
What should happen next?
Start Work Briefs, verification plans, handoff next steps, recommended checks, and workflow phase state turn context into a reviewable execution path. Where evaluated calibration exists, recommendations are ranked by observed outcome reliability.
Can I safely proceed?
Confidence profiles, outcome-weighted judgment, Safe Parallel Coding, collaboration guards, resource leases, release policy gates, stale warnings, and decision consistency checks expose weak evidence, overlap, explicit contradictions, or missing proof before risky actions.
What is available today
The current product answers the five-question model through concrete surfaces. Treat these as operational entry points and governed agent surfaces, not separate products.
Measured impact
The June 2026 hosted GPT-4.1 benchmark shows 6.3K selected tokens vs. a 32K raw-window baseline — 80% less context sent to the model, with higher answer quality on the same task. Public proof stays measurable and narrow; broader time-saved or ROI claims wait until workflow traces capture those measurements directly.
Inspect the controlled replayWorkflow Continuity
Start Work Briefs, What Changed For Me, PR Answer Packs, handoffs, phase commits, and resume context keep agent work from restarting at zero.
Context authority
Reviewed memory, provenance, source URLs, freshness, confidence, stale warnings, and validation state make context inspectable before agents trust it.
Context as Code
Context Control compares a reviewed ProjectContext manifest with tenant-scoped hosted documents, applies only approved add/update operations with remote hash preconditions, and emits detailed receipts without deleting unmanaged context.
Code impact
Code graph tools and companion impact commands give agents affected files, related tests, risk signals, and verification hints before implementation.
Why and decision capture
Why Capture covers confirmed workflow evidence, while GitHub PR and GitLab MR Rationale Packs extract explicit decisions, reasons, alternatives, and constraints from review text and commits. Every candidate keeps source provenance and remains pending human review.
Safe parallel coding
Team Sync, presence, leases, guard profiles, GitHub checks, and local companion commands reduce conflicting work across humans and agents.
Outcome-weighted judgment
Project Intelligence composes decisions, code graph context, outcome signals, advisor influence, and observed reliability into recommendations with confidence, evidence, counter-evidence, and caveats.
Hosted Outcome Intelligence
Projects can ingest, review, and aggregate Outcome Intelligence receipts by comparable task profile while excluding rejected evidence and keeping the results project-scoped.
Project policy and consistency
Decision Consistency evaluates reviewed project decisions, detects explicit contradictions, and produces review or blocking signals without allowing one project to influence another.
Controlled worker execution
Companion supports dry runs, explicit approval, bounded write scope, high-risk command blocking, acceptance criteria, and proof-required execution receipts.
Release and verification gates
Release policy gates, Project Policy decision-consistency receipts, verification plans, collaboration guard verdicts, and structured handoffs keep schema, auth, billing, deploy, and package blockers explicit before promotion.
Proof and replay assets
Agent Readiness Audit, ADE Adapter Packs, release gates, and Coding Intelligence Ledger exports turn the work loop into structured review and replay artifacts without dumping raw transcripts.
What comes next
These are net-new product surfaces that are not available today.
Standalone commit rationale
Extend the reviewed rationale flow to standalone commits that never pass through a pull request, merge request, or managed workflow, while preserving source provenance and human approval.
Cross-project outcome learning
Bring privacy-safe outcome patterns across comparable projects and teams while preserving tenant isolation and keeping every recommendation traceable to evidence.
Team policy management
Give teams one place to author policies, review exceptions, manage approvals, and govern shared safeguards across projects.
Hosted worker supervision
Supervise approved workers with budgets, visible progress, bounded permissions, and proof review from a hosted control surface.
Living project pages
Persist evidence-backed project pages that stay traceable to their sources, expose staleness clearly, and never replace canonical project knowledge.
How to use it in an agent workflow
- Connect the repository with create-snipara or Hosted MCP.
- Start risky or resumed work with snipara-companion brief or a hosted MCP context query.
- Inspect the Rationale Pack on each pull request or merge request, then review captured decision drafts before relying on them as current project decisions.
- Use code impact or symbol cards before changing routes, services, jobs, auth, billing, schema, deployment, or shared behavior.
- Let Project Policy decision-consistency receipts require review or block only when approved decision evidence matches the current action with enough confidence.
- Commit phases, hand off the session, and let the next agent resume from project-owned state instead of transcript memory.
What Snipara does not claim
- Snipara does not replace Claude Code, Codex, Cursor, ChatGPT, or your own agent runtime.
- Snipara does not silently launch or supervise worker agents. Controlled worker execution is explicit, approval-gated, and proof-required; broader hosted automation remains gated.
- Snipara does not treat every chat transcript as project memory.
- Snipara does not claim every project judgment is fully autonomous today.
- Judgment confidence is not a calibrated probability or advisor-grade certainty; it is an inspectable, sample-gated advisory signal.
- Project Intelligence and outcome-weighted judgment are the public category; memory is a component capability, not the category.
- Project Intelligence briefs are compiled evidence views, not canonical project truth.
- Reliability curves summarize observed outcomes for calibration; they are not causal proof.
- Outcome-weighted ranking applies only where evaluated calibration exists; otherwise Project Intelligence keeps static fallbacks and explicit caveats.
- Outcome Intelligence receipts are calibration evidence, not causal proof, canonical memory, global agent trust, or a Project Policy override.
- Human feedback on advisor receipts captures perishable pertinence; it is not an outcome signal or causal proof.
- A cold-start fuel status means matched decision or outcome evidence is thin, not that the system is unavailable.