Precia
How It WorksDevelopersChangelog
Login
Changelog

Precia changelog & release notes

Current public release v0.4.0

MCP governance, AI model routing, benchmark, web search, and scenario testing updates in one place.

Unreleased
May 16, 2026

MCP Tool Manager security draft

v0.4.1 draft

The next MCP update tightens tool visibility, approval behavior, and gateway safety before release.

MCPSecurityApprovals
4 highlights2 fixed

Highlights

4
  • Tool Manager lets admins enable discovered MCP tools and choose auto-execute or approval-required behavior.
  • Newly discovered tools start disabled, so chat only sees tools an admin has explicitly allowed.
  • The MCP gateway masks disabled tools from raw tool lists and blocks unknown or disabled tool calls.
  • Manual approval resume is wired into chat, with pending tool calls surfaced for review.

Fixes

2
  • Tool arguments now get size, schema, and simple injection-marker checks before execution.
  • Tool output is sanitized and marked as untrusted before it can be sent back to the model.
Latest shipped
May 13, 2026

MCP scenario workflows

v0.4.0

MCP testing moved into scenario setup with live capture, mock responses, and cleaner streaming behavior.

MCPScenariosModels
4 highlights2 fixed

Highlights

4
  • Scenarios can now include MCP workflows for testing tool-backed model behavior.
  • Live capture can populate mock responses from real MCP server requests.
  • Scenario setup now includes MCP archetypes and tool-validation controls.
  • OpenRouter models were refreshed so benchmark and chat choices use the latest available list.

Fixes

2
  • Streaming responses are more stable during tool-backed chat runs.
  • Long scenario text now wraps instead of overflowing the scenario view.
May 9, 2026

MCP scenarios and validation

v0.3.9

Scenario authoring gained MCP-specific setup tools, validation, and clearer scenario review surfaces.

MCPValidationScenarios
4 highlights2 fixed

Highlights

4
  • Preset scenarios can define MCP tool expectations and validation rules.
  • MCP archetypes make common tool-testing patterns faster to set up.
  • Scenario rows and review screens were updated for denser repeated editing.
  • Tool-call validation can compare expected calls, arguments, and response shape.

Fixes

2
  • MCP discovery and repository tests now cover more scenario setup paths.
  • Scenario view cleanup reduced noisy intermediate states while editing MCP cases.
May 3, 2026

Public MCP web search and model lists

v0.3.5

Web search moved onto a public MCP path and model picking became easier to scan.

SearchModelsWorkers
4 highlights2 fixed

Highlights

4
  • Precia web search can run through the public MCP web-search service.
  • Chat model lists were simplified into a compact one-line control.
  • The model picker popover was refreshed with clearer model metadata.
  • Provider keys can be forwarded server-side for request-scoped web search.

Fixes

2
  • Evaluation schema model changes now stay aligned with preset state.
  • Cloudflare Worker fetch calls avoid detached runtime fetch binding errors.
May 1, 2026

MCP control plane foundation

v0.3.0

Precia gained its first real MCP control plane, plus chat, theme, and search workflow updates.

MCPChatUI
4 highlights2 fixed

Highlights

4
  • MCP pages now cover servers, sessions, policies, approvals, and replay history.
  • MCP server setup validates safer URLs and supports connection testing, discovery, and audit trails.
  • EXA-backed web search is available as a provider option for new users.
  • Light theme, chat input, and chat navigation flows were refreshed.

Fixes

2
  • Chat submit and navigation interactions were cleaned up after the MCP shell work.
  • MCP connection and replay flows gained focused tests for safer follow-up changes.
April 3, 2026

Model Packs

v0.2.0

Presets now support pack-based model selection with per-preset customization.

Model PacksPresetsOpenRouter
3 highlights2 fixed

Highlights

3
  • Each preset now has its own Model Pack selector: Free, Recommended, Cheap, Balanced, Premium, Fast, and Structured.
  • You can fine-tune any pack with provider filters and manual model selection.
  • The Free pack uses zero-cost benchmark models, but judge model calls still cost money.

Fixes

2
  • OpenRouter privacy settings now surface blocked models before you apply changes.
  • Pack counts and saved selections now stay aligned when unavailable models are filtered out.
March 29, 2026

Preset Scenarios rework

v0.1.2

This update makes scenario editing cleaner and bulk actions easier to use.

ScenariosPresets
2 highlights1 fixed

Highlights

2
  • Preset Scenarios now uses cleaner card-based rows with inline controls.
  • Bulk select and bulk delete are now available from the sticky toolbar.

Fixes

1
  • Scenario selection now stays in sync as scenarios are added or removed.
March 29, 2026

Faster scenarios and cleaner preset flow

v0.1.1

This update speeds up setup and makes first-run flows easier to scan.

ScenariosOnboarding
3 highlights1 fixed

Highlights

3
  • Scenario generation now runs on a faster setup.
  • The sidebar opens by default on first visit.
  • New presets use names like "Preset #1".

Fixes

1
  • Scenario extraction now skips noisy duplicate assistant replies.
March 28, 2026

Benchmark and API foundation

March 2026

March work made Precia easier to trust, measure, and integrate before the later MCP releases.

BenchmarksOpenAPIJudging
4 highlights2 fixed

Highlights

4
  • Benchmark results gained richer cost, quality, duration, and model-ranking views.
  • Rubric and judge workflows were rebuilt around structured scoring and clearer scenario rules.
  • The API moved toward an OpenAPI source of truth with generated client hooks.
  • Web search, waitlist, analytics, and D1/R2 storage work landed as product foundations.

Fixes

2
  • Benchmark polling and chart behavior became more stable on long-running runs.
  • Database and migration work cleaned up limits, result storage, and production readiness.
© 2026 Precia. All rights reserved.
AppHow It WorksChangelog